classen3 commited on
Commit
a67be9a
1 Parent(s): a8e3541

Imported CS 441 audio/transcripts

Browse files
CS_441_2023_Spring_February_02,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78cc37a0ca7cfb6af05338a417c0f8df8b0c871987312e48c281db27acedf842
3
+ size 76791576
CS_441_2023_Spring_February_02,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_07,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d808b889b889801900685a4342b37fee7cf42cc60db1d6111aff0215203c747
3
+ size 76785048
CS_441_2023_Spring_February_07,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_09,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b83a68b45636e086c3866c52a1b83444f99063cbe016e1a603ef45f123c8ae2d
3
+ size 76790424
CS_441_2023_Spring_February_09,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_14,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f55c3315cebcc717120b207652aa22f0d60cde3776b4efc5afa674948b549118
3
+ size 76791576
CS_441_2023_Spring_February_14,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_16,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:23697f3f247a3b6d5ad45091f0716df72865f935d27dd3b945c026fdaec13a67
3
+ size 76789656
CS_441_2023_Spring_February_16,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_21,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a65e366399ca78085150b8d4f1a9e3c540459fbe7081c822b25b0ecec9b97b4
3
+ size 76789656
CS_441_2023_Spring_February_21,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_23,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:068ea9a11c749c07caebbaed875e0c66a7b01b8423296322087c23727cb4a92b
3
+ size 76790424
CS_441_2023_Spring_February_23,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_February_28,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6daad39d3f258680f1308a9d01713befc84e457936404723e960d4856172a6ed
3
+ size 76788504
CS_441_2023_Spring_February_28,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_January_17,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a8d6c4b6a10c378dd615eb87bee812b05561af0d10f2fcf1f30cdaf0f08a1f76
3
+ size 76794648
CS_441_2023_Spring_January_17,_2023.vtt ADDED
@@ -0,0 +1,4064 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ WEBVTT Kind: captions; Language: en-US
2
+
3
+ NOTE
4
+ Created on 2024-02-07T20:43:56.2906687Z by ClassTranscribe
5
+
6
+ 00:03:01.710 --> 00:03:04.290
7
+ Alright, good morning everybody.
8
+
9
+ 00:03:06.080 --> 00:03:07.190
10
+ It's good to see you all.
11
+
12
+ 00:03:07.190 --> 00:03:10.625
13
+ Hope you had a good break, so I'm going
14
+
15
+ 00:03:10.625 --> 00:03:11.430
16
+ to get started.
17
+
18
+ 00:03:12.600 --> 00:03:15.170
19
+ So this is CS441 Applied machine
20
+
21
+ 00:03:15.170 --> 00:03:17.340
22
+ learning and I'm Derek Hoiem, the
23
+
24
+ 00:03:17.340 --> 00:03:17.930
25
+ Instructor.
26
+
27
+ 00:03:19.320 --> 00:03:21.240
28
+ So today I'm going to just tell you a
29
+
30
+ 00:03:21.240 --> 00:03:22.510
31
+ little bit about myself.
32
+
33
+ 00:03:22.510 --> 00:03:24.120
34
+ I'll talk about machine learning,
35
+
36
+ 00:03:24.120 --> 00:03:27.030
37
+ applied machine learning and a bit
38
+
39
+ 00:03:27.030 --> 00:03:28.530
40
+ about the course.
41
+
42
+ 00:03:28.530 --> 00:03:30.616
43
+ I'll give an outline in the course and
44
+
45
+ 00:03:30.616 --> 00:03:32.376
46
+ talk about some of the logistics of the
47
+
48
+ 00:03:32.376 --> 00:03:32.579
49
+ course.
50
+
51
+ 00:03:33.190 --> 00:03:34.510
52
+ And we'll probably end a little bit
53
+
54
+ 00:03:34.510 --> 00:03:36.480
55
+ early so that there's time for you to
56
+
57
+ 00:03:36.480 --> 00:03:37.840
58
+ ask any questions that you have about
59
+
60
+ 00:03:37.840 --> 00:03:41.390
61
+ the course, either individually or you
62
+
63
+ 00:03:41.390 --> 00:03:44.247
64
+ can ask them at the ask them in general
65
+
66
+ 00:03:44.247 --> 00:03:45.880
67
+ at the end, at the end of class.
68
+
69
+ 00:03:47.860 --> 00:03:50.440
70
+ So first a little bit about me.
71
+
72
+ 00:03:50.440 --> 00:03:51.670
73
+ You might know some of this if you've
74
+
75
+ 00:03:51.670 --> 00:03:53.890
76
+ taken a class with me before, but I was
77
+
78
+ 00:03:53.890 --> 00:03:55.800
79
+ raised in upstate New York, right where
80
+
81
+ 00:03:55.800 --> 00:03:57.240
82
+ the circle is and the Hudson River
83
+
84
+ 00:03:57.240 --> 00:03:57.630
85
+ valley.
86
+
87
+ 00:03:57.630 --> 00:03:59.320
88
+ This is the lake that I grew up on.
89
+
90
+ 00:04:00.610 --> 00:04:03.730
91
+ I actually was interested in machine
92
+
93
+ 00:04:03.730 --> 00:04:06.380
94
+ learning and AI for quite a long time.
95
+
96
+ 00:04:07.750 --> 00:04:09.650
97
+ When I decided to go when I went to
98
+
99
+ 00:04:09.650 --> 00:04:11.730
100
+ undergrad at Buffalo, I decided to
101
+
102
+ 00:04:11.730 --> 00:04:14.280
103
+ major in electrical engineering because
104
+
105
+ 00:04:14.280 --> 00:04:15.420
106
+ I.
107
+
108
+ 00:04:15.820 --> 00:04:17.565
109
+ Because it has a good strong background
110
+
111
+ 00:04:17.565 --> 00:04:20.070
112
+ in math, and I was advised that it's a
113
+
114
+ 00:04:20.070 --> 00:04:21.540
115
+ good foundation for anything else I
116
+
117
+ 00:04:21.540 --> 00:04:22.430
118
+ wanted to do.
119
+
120
+ 00:04:22.430 --> 00:04:25.170
121
+ But my roommate was taking computer
122
+
123
+ 00:04:25.170 --> 00:04:27.490
124
+ science and I liked his Assignments,
125
+
126
+ 00:04:27.490 --> 00:04:29.050
127
+ and I started just doing them for fun
128
+
129
+ 00:04:29.050 --> 00:04:31.610
130
+ and started taking more CS courses.
131
+
132
+ 00:04:31.610 --> 00:04:33.360
133
+ And I kind of made a beeline in my
134
+
135
+ 00:04:33.360 --> 00:04:36.239
136
+ curriculum for AI and machine learning,
137
+
138
+ 00:04:36.240 --> 00:04:38.020
139
+ which we're not nearly as popular.
140
+
141
+ 00:04:38.020 --> 00:04:39.830
142
+ At the time I was there was only one
143
+
144
+ 00:04:39.830 --> 00:04:41.435
145
+ graduate course in machine learning,
146
+
147
+ 00:04:41.435 --> 00:04:43.860
148
+ and I was the only undergraduate who
149
+
150
+ 00:04:43.860 --> 00:04:45.940
151
+ took that course that year.
152
+
153
+ 00:04:47.650 --> 00:04:50.828
154
+ And so I ended up adding on the
155
+
156
+ 00:04:50.828 --> 00:04:52.880
157
+ computer the computer engineering
158
+
159
+ 00:04:52.880 --> 00:04:54.240
160
+ degree as well, just because I had
161
+
162
+ 00:04:54.240 --> 00:04:57.340
163
+ taken so many CS courses that I was
164
+
165
+ 00:04:57.340 --> 00:04:58.890
166
+ just able to do that.
167
+
168
+ 00:05:00.000 --> 00:05:02.010
169
+ When I after I graduated, I went to
170
+
171
+ 00:05:02.010 --> 00:05:03.570
172
+ Carnegie Mellon and I went for
173
+
174
+ 00:05:03.570 --> 00:05:04.410
175
+ Robotics.
176
+
177
+ 00:05:04.600 --> 00:05:08.260
178
+ And I started working with somebody
179
+
180
+ 00:05:08.260 --> 00:05:10.866
181
+ named Henry Schneiderman, who made at
182
+
183
+ 00:05:10.866 --> 00:05:12.500
184
+ the what was at the time the most
185
+
186
+ 00:05:12.500 --> 00:05:15.580
187
+ accurate face detector, and then.
188
+
189
+ 00:05:15.660 --> 00:05:19.260
190
+ After he left to start a company and
191
+
192
+ 00:05:19.260 --> 00:05:21.120
193
+ then I started working with Elisha
194
+
195
+ 00:05:21.120 --> 00:05:24.810
196
+ Frozen Marshalli Bear doing applying
197
+
198
+ 00:05:24.810 --> 00:05:26.770
199
+ machine learning to try to recognize
200
+
201
+ 00:05:26.770 --> 00:05:29.430
202
+ geometry to create 3D models based on
203
+
204
+ 00:05:29.430 --> 00:05:31.210
205
+ single view recognition.
206
+
207
+ 00:05:32.140 --> 00:05:35.175
208
+ Then, after my PhD, I came here.
209
+
210
+ 00:05:35.175 --> 00:05:38.280
211
+ I did like a postdoc fellowship for a
212
+
213
+ 00:05:38.280 --> 00:05:39.360
214
+ little bit, and then I've been a
215
+
216
+ 00:05:39.360 --> 00:05:40.720
217
+ professor here ever since.
218
+
219
+ 00:05:43.140 --> 00:05:44.780
220
+ So I'll tell you a little bit about my
221
+
222
+ 00:05:44.780 --> 00:05:45.530
223
+ research.
224
+
225
+ 00:05:45.530 --> 00:05:47.870
226
+ Actually the first two projects that I
227
+
228
+ 00:05:47.870 --> 00:05:51.480
229
+ did were on music identification and
230
+
231
+ 00:05:51.480 --> 00:05:54.440
232
+ sound detection and then this is the.
233
+
234
+ 00:05:54.440 --> 00:05:57.460
235
+ And then I did another one on object on
236
+
237
+ 00:05:57.460 --> 00:05:59.820
238
+ retrieving images based on objects in
239
+
240
+ 00:05:59.820 --> 00:06:00.050
241
+ them.
242
+
243
+ 00:06:00.620 --> 00:06:02.440
244
+ And then this was like my first main
245
+
246
+ 00:06:02.440 --> 00:06:04.970
247
+ project which was part of my thesis.
248
+
249
+ 00:06:04.970 --> 00:06:08.310
250
+ So at the time, if you wanted to create
251
+
252
+ 00:06:08.310 --> 00:06:10.760
253
+ a 3D model of a scene from images, you
254
+
255
+ 00:06:10.760 --> 00:06:12.710
256
+ basically solved a bunch of algebraic
257
+
258
+ 00:06:12.710 --> 00:06:13.460
259
+ constraints.
260
+
261
+ 00:06:14.160 --> 00:06:15.860
262
+ Based on correspondences between the
263
+
264
+ 00:06:15.860 --> 00:06:16.630
265
+ images.
266
+
267
+ 00:06:16.630 --> 00:06:19.680
268
+ But we had this idea that since people
269
+
270
+ 00:06:19.680 --> 00:06:22.010
271
+ are able to see an image like this one
272
+
273
+ 00:06:22.010 --> 00:06:24.823
274
+ and understand the 3D scene behind the
275
+
276
+ 00:06:24.823 --> 00:06:26.650
277
+ image, maybe we can use machine
278
+
279
+ 00:06:26.650 --> 00:06:29.090
280
+ learning to recognize the surfaces and
281
+
282
+ 00:06:29.090 --> 00:06:31.540
283
+ the image and then use that to create a
284
+
285
+ 00:06:31.540 --> 00:06:32.650
286
+ simple 3D model.
287
+
288
+ 00:06:35.370 --> 00:06:37.410
289
+ Currently a couple of the main
290
+
291
+ 00:06:37.410 --> 00:06:40.625
292
+ directions I work in are one of them is
293
+
294
+ 00:06:40.625 --> 00:06:42.070
295
+ this thing called neural radiance
296
+
297
+ 00:06:42.070 --> 00:06:44.610
298
+ fields, and basically the idea is to
299
+
300
+ 00:06:44.610 --> 00:06:47.360
301
+ use the machine learning system as a
302
+
303
+ 00:06:47.360 --> 00:06:50.000
304
+ kind of compression to encode all the
305
+
306
+ 00:06:50.000 --> 00:06:51.905
307
+ geometry and appearance information in
308
+
309
+ 00:06:51.905 --> 00:06:54.525
310
+ the scene so that you can render out
311
+
312
+ 00:06:54.525 --> 00:06:57.895
313
+ the surface normals or the depth or the
314
+
315
+ 00:06:57.895 --> 00:06:59.940
316
+ appearance of images from arbitrary
317
+
318
+ 00:06:59.940 --> 00:07:00.277
319
+ viewpoints.
320
+
321
+ 00:07:00.277 --> 00:07:02.973
322
+ And so this is one of the directions
323
+
324
+ 00:07:02.973 --> 00:07:03.830
325
+ that we're working.
326
+
327
+ 00:07:04.500 --> 00:07:06.470
328
+ And another one is what I call general
329
+
330
+ 00:07:06.470 --> 00:07:07.650
331
+ purpose Learning.
332
+
333
+ 00:07:07.960 --> 00:07:10.930
334
+ And as you'll see, in a lot of machine
335
+
336
+ 00:07:10.930 --> 00:07:13.190
337
+ learning, the goal is to solve one
338
+
339
+ 00:07:13.190 --> 00:07:16.620
340
+ particular kind of Prediction task, or
341
+
342
+ 00:07:16.620 --> 00:07:20.120
343
+ like unsupervised learning task.
344
+
345
+ 00:07:20.120 --> 00:07:22.525
346
+ But people were quite differently.
347
+
348
+ 00:07:22.525 --> 00:07:25.250
349
+ We don't have like a single classifier
350
+
351
+ 00:07:25.250 --> 00:07:27.905
352
+ in our head or a single purpose that
353
+
354
+ 00:07:27.905 --> 00:07:29.900
355
+ our mind is designed for.
356
+
357
+ 00:07:29.900 --> 00:07:32.916
358
+ Instead, we have the ability to sense a
359
+
360
+ 00:07:32.916 --> 00:07:34.638
361
+ lot of things and then we can do a lot
362
+
363
+ 00:07:34.638 --> 00:07:36.650
364
+ of things with our body and our speech,
365
+
366
+ 00:07:36.650 --> 00:07:38.530
367
+ and so we're able to solve kind of an
368
+
369
+ 00:07:38.530 --> 00:07:39.310
370
+ infinite range.
371
+
372
+ 00:07:39.400 --> 00:07:43.100
373
+ Tasks that are senses and our actions
374
+
375
+ 00:07:43.100 --> 00:07:44.450
376
+ allow us to perform.
377
+
378
+ 00:07:44.450 --> 00:07:46.140
379
+ And so we're trying to do the same
380
+
381
+ 00:07:46.140 --> 00:07:47.790
382
+ thing for machine learning, to create
383
+
384
+ 00:07:47.790 --> 00:07:50.493
385
+ systems that are able to receive some
386
+
387
+ 00:07:50.493 --> 00:07:53.538
388
+ kinds of inputs, like text and images,
389
+
390
+ 00:07:53.538 --> 00:07:56.330
391
+ and produce outputs which could also be
392
+
393
+ 00:07:56.330 --> 00:07:57.457
394
+ like text or images.
395
+
396
+ 00:07:57.457 --> 00:08:00.180
397
+ And then do any kinds of tasks that
398
+
399
+ 00:08:00.180 --> 00:08:02.040
400
+ fall within that range of the input and
401
+
402
+ 00:08:02.040 --> 00:08:02.833
403
+ output modality.
404
+
405
+ 00:08:02.833 --> 00:08:05.240
406
+ And then we also work on ways to try to
407
+
408
+ 00:08:05.240 --> 00:08:08.380
409
+ extend these systems, for example in
410
+
411
+ 00:08:08.380 --> 00:08:08.750
412
+ this.
413
+
414
+ 00:08:09.410 --> 00:08:11.260
415
+ And this slide here is showing that we
416
+
417
+ 00:08:11.260 --> 00:08:14.000
418
+ created this system that is able to map
419
+
420
+ 00:08:14.000 --> 00:08:16.420
421
+ from images in a text prompt into some
422
+
423
+ 00:08:16.420 --> 00:08:17.320
424
+ kind of answer.
425
+
426
+ 00:08:18.080 --> 00:08:20.260
427
+ And then the system is able to learn
428
+
429
+ 00:08:20.260 --> 00:08:23.470
430
+ from web images to learn new concepts.
431
+
432
+ 00:08:23.470 --> 00:08:25.550
433
+ So we were able to, we provided it with
434
+
435
+ 00:08:25.550 --> 00:08:27.682
436
+ like a set of keywords about COVID, and
437
+
438
+ 00:08:27.682 --> 00:08:29.760
439
+ then it was able to just download a
440
+
441
+ 00:08:29.760 --> 00:08:31.863
442
+ bunch of images from Bing and then
443
+
444
+ 00:08:31.863 --> 00:08:33.530
445
+ learn how to answer questions and
446
+
447
+ 00:08:33.530 --> 00:08:35.300
448
+ caption and detect things that are
449
+
450
+ 00:08:35.300 --> 00:08:37.290
451
+ related to COVID, which we're not in
452
+
453
+ 00:08:37.290 --> 00:08:38.795
454
+ any of the original data that I trained
455
+
456
+ 00:08:38.795 --> 00:08:39.050
457
+ in.
458
+
459
+ 00:08:42.460 --> 00:08:44.460
460
+ I've also used machine learning in a
461
+
462
+ 00:08:44.460 --> 00:08:47.414
463
+ lot of other ways, so let me just.
464
+
465
+ 00:08:47.414 --> 00:08:49.045
466
+ I just want to make sure I think the
467
+
468
+ 00:08:49.045 --> 00:08:50.650
469
+ mic's working, but I just want to.
470
+
471
+ 00:08:51.720 --> 00:08:54.230
472
+ Make sure that it's OK.
473
+
474
+ 00:08:54.230 --> 00:08:56.480
475
+ It'll pick up a little better down here
476
+
477
+ 00:08:56.480 --> 00:08:58.370
478
+ because for the recording.
479
+
480
+ 00:09:00.030 --> 00:09:02.375
481
+ So I've used machine learning in lots
482
+
483
+ 00:09:02.375 --> 00:09:02.960
484
+ of different ways.
485
+
486
+ 00:09:02.960 --> 00:09:05.540
487
+ In my research I've done things like
488
+
489
+ 00:09:05.540 --> 00:09:07.515
490
+ object detection, image classification,
491
+
492
+ 00:09:07.515 --> 00:09:10.020
493
+ 3D scene modeling, Generating
494
+
495
+ 00:09:10.020 --> 00:09:12.300
496
+ animations, visual question answering,
497
+
498
+ 00:09:12.300 --> 00:09:14.240
499
+ phrase grounding, sound detection.
500
+
501
+ 00:09:14.240 --> 00:09:16.690
502
+ So lots of different, lots of different
503
+
504
+ 00:09:16.690 --> 00:09:17.390
505
+ use cases.
506
+
507
+ 00:09:18.850 --> 00:09:20.566
508
+ And then I've also used machine
509
+
510
+ 00:09:20.566 --> 00:09:22.700
511
+ learning in Application.
512
+
513
+ 00:09:22.700 --> 00:09:25.240
514
+ So I cofounded this company,
515
+
516
+ 00:09:25.240 --> 00:09:27.420
517
+ Reconstruct, where we take images of a
518
+
519
+ 00:09:27.420 --> 00:09:29.460
520
+ construction site and bring it together
521
+
522
+ 00:09:29.460 --> 00:09:31.525
523
+ with building plans and schedule to
524
+
525
+ 00:09:31.525 --> 00:09:32.955
526
+ help all the stakeholders understand
527
+
528
+ 00:09:32.955 --> 00:09:35.480
529
+ the progress and whether things are
530
+
531
+ 00:09:35.480 --> 00:09:37.230
532
+ being built according to the plan.
533
+
534
+ 00:09:37.230 --> 00:09:40.694
535
+ And part of this, some of the 3D
536
+
537
+ 00:09:40.694 --> 00:09:42.020
538
+ construction doesn't use machine
539
+
540
+ 00:09:42.020 --> 00:09:43.990
541
+ learning, but some of it does to help
542
+
543
+ 00:09:43.990 --> 00:09:47.170
544
+ create more complete Models and then to
545
+
546
+ 00:09:47.170 --> 00:09:48.880
547
+ recognize things in the scene.
548
+
549
+ 00:09:48.950 --> 00:09:51.060
550
+ To help monitor their progress.
551
+
552
+ 00:09:53.330 --> 00:09:56.100
553
+ So all of that is to say that I've had
554
+
555
+ 00:09:56.100 --> 00:09:58.100
556
+ quite a lot of experience using machine
557
+
558
+ 00:09:58.100 --> 00:09:59.380
559
+ learning in a variety of different
560
+
561
+ 00:09:59.380 --> 00:10:02.205
562
+ ways, many different modalities, many
563
+
564
+ 00:10:02.205 --> 00:10:04.940
565
+ different kinds of applications, both
566
+
567
+ 00:10:04.940 --> 00:10:08.220
568
+ research and in commercial types of
569
+
570
+ 00:10:08.220 --> 00:10:08.620
571
+ applications.
572
+
573
+ 00:10:10.360 --> 00:10:12.267
574
+ So what is machine learning?
575
+
576
+ 00:10:12.267 --> 00:10:14.170
577
+ So everyone has their own definition.
578
+
579
+ 00:10:14.170 --> 00:10:16.740
580
+ But the way that I think about it is
581
+
582
+ 00:10:16.740 --> 00:10:19.195
583
+ that it's machine learning spins Raw
584
+
585
+ 00:10:19.195 --> 00:10:21.710
586
+ data into gold, so it's pretty easy to
587
+
588
+ 00:10:21.710 --> 00:10:22.180
589
+ get data.
590
+
591
+ 00:10:22.180 --> 00:10:23.880
592
+ Now there's all kinds of data.
593
+
594
+ 00:10:23.880 --> 00:10:26.210
595
+ Whenever you use applications, you're
596
+
597
+ 00:10:26.210 --> 00:10:29.500
598
+ providing those Application providers
599
+
600
+ 00:10:29.500 --> 00:10:30.900
601
+ with your data.
602
+
603
+ 00:10:30.900 --> 00:10:33.180
604
+ A lot of times you can download lots of
605
+
606
+ 00:10:33.180 --> 00:10:34.470
607
+ information on the Internet.
608
+
609
+ 00:10:34.470 --> 00:10:36.850
610
+ There's just data everywhere, but by
611
+
612
+ 00:10:36.850 --> 00:10:38.970
613
+ itself, that data is not very useful.
614
+
615
+ 00:10:39.030 --> 00:10:41.270
616
+ It's too much to manually go through.
617
+
618
+ 00:10:41.270 --> 00:10:42.610
619
+ It's not something that you can
620
+
621
+ 00:10:42.610 --> 00:10:46.320
622
+ actually solve a problem with, and so
623
+
624
+ 00:10:46.320 --> 00:10:50.460
625
+ machine learning is really the ability
626
+
627
+ 00:10:50.460 --> 00:10:52.610
628
+ to take all of that Raw data and turn
629
+
630
+ 00:10:52.610 --> 00:10:55.066
631
+ it into something useful to be able to
632
+
633
+ 00:10:55.066 --> 00:10:57.390
634
+ create predictive models or to create
635
+
636
+ 00:10:57.390 --> 00:10:58.080
637
+ insights.
638
+
639
+ 00:10:58.710 --> 00:11:01.760
640
+ So some examples are here for the Alexa
641
+
642
+ 00:11:01.760 --> 00:11:04.060
643
+ speech recognition where you.
644
+
645
+ 00:11:04.340 --> 00:11:06.940
646
+ Where they learn from audio
647
+
648
+ 00:11:06.940 --> 00:11:09.920
649
+ transcriptions to be able to take the
650
+
651
+ 00:11:09.920 --> 00:11:13.465
652
+ voice information that you speak into
653
+
654
+ 00:11:13.465 --> 00:11:16.235
655
+ the Alexa app and turn it into text and
656
+
657
+ 00:11:16.235 --> 00:11:18.390
658
+ then it's then turned into other
659
+
660
+ 00:11:18.390 --> 00:11:19.310
661
+ information.
662
+
663
+ 00:11:20.450 --> 00:11:23.110
664
+ Or product recommendations?
665
+
666
+ 00:11:23.110 --> 00:11:24.860
667
+ Autonomous vehicles?
668
+
669
+ 00:11:26.330 --> 00:11:30.186
670
+ Text generation like GPT 3 or GPT chat
671
+
672
+ 00:11:30.186 --> 00:11:32.230
673
+ or ChatGPT image generation.
674
+
675
+ 00:11:32.230 --> 00:11:34.675
676
+ And then there's a link there that you
677
+
678
+ 00:11:34.675 --> 00:11:35.580
679
+ can check out later.
680
+
681
+ 00:11:35.580 --> 00:11:37.780
682
+ It's a data visualization of Twitter
683
+
684
+ 00:11:37.780 --> 00:11:40.780
685
+ where trying to take like all the tons
686
+
687
+ 00:11:40.780 --> 00:11:42.400
688
+ of information about different tweets
689
+
690
+ 00:11:42.400 --> 00:11:44.642
691
+ and organize it in a way so that you
692
+
693
+ 00:11:44.642 --> 00:11:47.240
694
+ can draw insights about like social
695
+
696
+ 00:11:47.240 --> 00:11:49.780
697
+ trends and kind of what's going on in
698
+
699
+ 00:11:49.780 --> 00:11:50.630
700
+ different places.
701
+
702
+ 00:11:54.780 --> 00:11:58.300
703
+ So we tend to think in classes.
704
+
705
+ 00:11:58.300 --> 00:12:00.710
706
+ Especially we tend to think about
707
+
708
+ 00:12:00.710 --> 00:12:03.250
709
+ machine learning as a problem of
710
+
711
+ 00:12:03.250 --> 00:12:09.216
712
+ designing algorithms that will map your
713
+
714
+ 00:12:09.216 --> 00:12:11.900
715
+ that allow you to learn from the
716
+
717
+ 00:12:11.900 --> 00:12:14.410
718
+ training data and to achieve good test
719
+
720
+ 00:12:14.410 --> 00:12:16.880
721
+ performance in your Prediction task
722
+
723
+ 00:12:16.880 --> 00:12:18.150
724
+ according to the test data.
725
+
726
+ 00:12:18.940 --> 00:12:20.920
727
+ But in the real world, there's actually
728
+
729
+ 00:12:20.920 --> 00:12:22.310
730
+ a bigger problem.
731
+
732
+ 00:12:22.310 --> 00:12:24.646
733
+ There's a bigger infrastructure around
734
+
735
+ 00:12:24.646 --> 00:12:25.518
736
+ machine learning.
737
+
738
+ 00:12:25.518 --> 00:12:27.945
739
+ And so while we're going to focus on
740
+
741
+ 00:12:27.945 --> 00:12:30.240
742
+ the Algorithm and model development,
743
+
744
+ 00:12:30.240 --> 00:12:32.099
745
+ it's also important to be aware of the
746
+
747
+ 00:12:32.100 --> 00:12:34.154
748
+ broader context of machine learning.
749
+
750
+ 00:12:34.154 --> 00:12:36.800
751
+ So if you were to ever go work for a
752
+
753
+ 00:12:36.800 --> 00:12:39.465
754
+ company and they say that they want you
755
+
756
+ 00:12:39.465 --> 00:12:43.100
757
+ to build some kind of predictor or some
758
+
759
+ 00:12:43.100 --> 00:12:45.583
760
+ data cluster or something like that,
761
+
762
+ 00:12:45.583 --> 00:12:48.230
763
+ you would actually go through multiple
764
+
765
+ 00:12:48.230 --> 00:12:49.590
766
+ different phases, so the 1st.
767
+
768
+ 00:12:50.360 --> 00:12:52.445
769
+ And potentially the most important is
770
+
771
+ 00:12:52.445 --> 00:12:54.710
772
+ the data preparation where you collect
773
+
774
+ 00:12:54.710 --> 00:12:56.326
775
+ and curate the data.
776
+
777
+ 00:12:56.326 --> 00:12:59.147
778
+ So you have to find examples you have.
779
+
780
+ 00:12:59.147 --> 00:13:00.970
781
+ You may need to Annotate it or hire
782
+
783
+ 00:13:00.970 --> 00:13:03.370
784
+ annotators or create annotation tools.
785
+
786
+ 00:13:03.370 --> 00:13:05.858
787
+ And then you split your data often into
788
+
789
+ 00:13:05.858 --> 00:13:07.740
790
+ a training set, validation set and a
791
+
792
+ 00:13:07.740 --> 00:13:08.278
793
+ test set.
794
+
795
+ 00:13:08.278 --> 00:13:10.350
796
+ So the training set is what you would
797
+
798
+ 00:13:10.350 --> 00:13:11.938
799
+ use to tune your Models, the validation
800
+
801
+ 00:13:11.938 --> 00:13:14.441
802
+ is what you use to select your Models,
803
+
804
+ 00:13:14.441 --> 00:13:17.685
805
+ and the test set is for your final
806
+
807
+ 00:13:17.685 --> 00:13:20.210
808
+ Final like estimate of your systems.
809
+
810
+ 00:13:20.270 --> 00:13:21.060
811
+ Performance.
812
+
813
+ 00:13:22.370 --> 00:13:23.920
814
+ Then once you have that, then you can
815
+
816
+ 00:13:23.920 --> 00:13:25.115
817
+ develop your algorithm.
818
+
819
+ 00:13:25.115 --> 00:13:27.560
820
+ You can decide what kind of machine
821
+
822
+ 00:13:27.560 --> 00:13:29.010
823
+ learning algorithm you're going to use.
824
+
825
+ 00:13:29.010 --> 00:13:30.560
826
+ What are the hyperparameters.
827
+
828
+ 00:13:31.570 --> 00:13:33.520
829
+ What kinds of objectives you're going
830
+
831
+ 00:13:33.520 --> 00:13:37.120
832
+ to give to your Algorithm and you train
833
+
834
+ 00:13:37.120 --> 00:13:38.850
835
+ it and evaluate it?
836
+
837
+ 00:13:38.850 --> 00:13:41.350
838
+ Often that process takes quite a lot of
839
+
840
+ 00:13:41.350 --> 00:13:44.193
841
+ time and it may lead you to go back to
842
+
843
+ 00:13:44.193 --> 00:13:46.050
844
+ the data preparation stage if you
845
+
846
+ 00:13:46.050 --> 00:13:47.400
847
+ realize that you're not going to be
848
+
849
+ 00:13:47.400 --> 00:13:49.960
850
+ able to reach the applications
851
+
852
+ 00:13:49.960 --> 00:13:51.030
853
+ performance requirements.
854
+
855
+ 00:13:52.350 --> 00:13:54.100
856
+ Once you're finally feel like you've
857
+
858
+ 00:13:54.100 --> 00:13:55.922
859
+ finished developing the Algorithm, then
860
+
861
+ 00:13:55.922 --> 00:13:58.430
862
+ you would test it on a test set, which
863
+
864
+ 00:13:58.430 --> 00:14:00.230
865
+ ideally you haven't used at any stage
866
+
867
+ 00:14:00.230 --> 00:14:01.520
868
+ before, so that it gives you an
869
+
870
+ 00:14:01.520 --> 00:14:03.480
871
+ unbiased estimate of your performance,
872
+
873
+ 00:14:03.480 --> 00:14:05.640
874
+ and then finally you integrate it into
875
+
876
+ 00:14:05.640 --> 00:14:06.570
877
+ your Application.
878
+
879
+ 00:14:07.210 --> 00:14:08.600
880
+ And.
881
+
882
+ 00:14:08.710 --> 00:14:11.160
883
+ And usually this Prediction engine that
884
+
885
+ 00:14:11.160 --> 00:14:13.170
886
+ you've built will only be one part of
887
+
888
+ 00:14:13.170 --> 00:14:15.410
889
+ an entire solution, and so it needs to
890
+
891
+ 00:14:15.410 --> 00:14:16.853
892
+ work well with the rest of that
893
+
894
+ 00:14:16.853 --> 00:14:17.139
895
+ solution.
896
+
897
+ 00:14:18.420 --> 00:14:22.480
898
+ So as an example, consider the voice
899
+
900
+ 00:14:22.480 --> 00:14:24.685
901
+ recognition and Alexa.
902
+
903
+ 00:14:24.685 --> 00:14:27.896
904
+ So they had a really challenging
905
+
906
+ 00:14:27.896 --> 00:14:30.540
907
+ problem that probably many of you have
908
+
909
+ 00:14:30.540 --> 00:14:34.120
910
+ at least used Alexa app sometime.
911
+
912
+ 00:14:34.120 --> 00:14:35.696
913
+ But there's a really challenging
914
+
915
+ 00:14:35.696 --> 00:14:37.370
916
+ problem that you've got like some disc
917
+
918
+ 00:14:37.370 --> 00:14:40.280
919
+ with some microphones on it and it you
920
+
921
+ 00:14:40.280 --> 00:14:43.945
922
+ need the disk needs to be able to
923
+
924
+ 00:14:43.945 --> 00:14:45.855
925
+ interpret your speech and it needs to
926
+
927
+ 00:14:45.855 --> 00:14:47.690
928
+ be able to interpret speech for a wide
929
+
930
+ 00:14:47.690 --> 00:14:48.850
931
+ range of people.
932
+
933
+ 00:14:49.140 --> 00:14:51.430
934
+ That are not just like talking into it.
935
+
936
+ 00:14:51.430 --> 00:14:53.550
937
+ You might talk into your cell phone,
938
+
939
+ 00:14:53.550 --> 00:14:55.425
940
+ but they could be shouting from across
941
+
942
+ 00:14:55.425 --> 00:14:56.190
943
+ the house.
944
+
945
+ 00:14:56.190 --> 00:14:59.710
946
+ Alexa, turn on turn play The Beatles or
947
+
948
+ 00:14:59.710 --> 00:15:00.870
949
+ Alexa, what's the temperature?
950
+
951
+ 00:15:01.490 --> 00:15:05.090
952
+ And Alexa needs to turn that into text
953
+
954
+ 00:15:05.090 --> 00:15:08.670
955
+ and then use other engines in order to
956
+
957
+ 00:15:08.670 --> 00:15:09.880
958
+ produce an answer that might be
959
+
960
+ 00:15:09.880 --> 00:15:10.980
961
+ relevant question.
962
+
963
+ 00:15:12.790 --> 00:15:13.276
964
+ Yeah.
965
+
966
+ 00:15:13.276 --> 00:15:15.530
967
+ Chat, ChatGPT is another example, but
968
+
969
+ 00:15:15.530 --> 00:15:17.040
970
+ that's quite different, yeah.
971
+
972
+ 00:15:18.220 --> 00:15:23.290
973
+ And so for you could Alexa could turn
974
+
975
+ 00:15:23.290 --> 00:15:23.540
976
+ that.
977
+
978
+ 00:15:23.540 --> 00:15:25.500
979
+ I'm sure that there are ChatGPT apps
980
+
981
+ 00:15:25.500 --> 00:15:26.980
982
+ that are connected to Alexa already
983
+
984
+ 00:15:26.980 --> 00:15:29.610
985
+ where you can then say, Alexa write my
986
+
987
+ 00:15:29.610 --> 00:15:30.650
988
+ essay for me.
989
+
990
+ 00:15:35.280 --> 00:15:36.540
991
+ So.
992
+
993
+ 00:15:37.030 --> 00:15:38.660
994
+ So for so they had this really
995
+
996
+ 00:15:38.660 --> 00:15:40.030
997
+ challenging speech recognition
998
+
999
+ 00:15:40.030 --> 00:15:41.846
1000
+ algorithm and they realized that the
1001
+
1002
+ 00:15:41.846 --> 00:15:43.300
1003
+ state-of-the-art wasn't up to the task.
1004
+
1005
+ 00:15:43.300 --> 00:15:46.400
1006
+ So they needed to achieve a certain
1007
+
1008
+ 00:15:46.400 --> 00:15:48.853
1009
+ word error rate, which means that the
1010
+
1011
+ 00:15:48.853 --> 00:15:51.650
1012
+ rate, the error rate for converting the
1013
+
1014
+ 00:15:51.650 --> 00:15:54.210
1015
+ spoken words into text.
1016
+
1017
+ 00:15:55.090 --> 00:15:57.053
1018
+ And they were kind of far below that.
1019
+
1020
+ 00:15:57.053 --> 00:15:59.120
1021
+ And so first, they started buying
1022
+
1023
+ 00:15:59.120 --> 00:16:01.590
1024
+ companies that were developing speech
1025
+
1026
+ 00:16:01.590 --> 00:16:02.656
1027
+ recognition engines.
1028
+
1029
+ 00:16:02.656 --> 00:16:05.280
1030
+ They started using deep learning.
1031
+
1032
+ 00:16:05.280 --> 00:16:06.930
1033
+ They started to try to improve their
1034
+
1035
+ 00:16:06.930 --> 00:16:10.002
1036
+ Algorithms, collect data, Annotate the
1037
+
1038
+ 00:16:10.002 --> 00:16:10.445
1039
+ data.
1040
+
1041
+ 00:16:10.445 --> 00:16:13.360
1042
+ And they had meetings with Jeff Bezos
1043
+
1044
+ 00:16:13.360 --> 00:16:18.590
1045
+ and said, we expect that within 10
1046
+
1047
+ 00:16:18.590 --> 00:16:20.120
1048
+ years, we're going to achieve the word
1049
+
1050
+ 00:16:20.120 --> 00:16:21.470
1051
+ error rate that we need.
1052
+
1053
+ 00:16:21.470 --> 00:16:23.850
1054
+ And basically Jeff Bezos, this was like
1055
+
1056
+ 00:16:23.850 --> 00:16:25.370
1057
+ his favorite project and he's like,
1058
+
1059
+ 00:16:25.370 --> 00:16:25.580
1060
+ well.
1061
+
1062
+ 00:16:25.630 --> 00:16:27.130
1063
+ You guys aren't thinking big enough.
1064
+
1065
+ 00:16:27.130 --> 00:16:29.220
1066
+ Basically we have tons of money.
1067
+
1068
+ 00:16:29.220 --> 00:16:31.380
1069
+ Figure out how you can do it faster.
1070
+
1071
+ 00:16:31.380 --> 00:16:33.390
1072
+ And So what they ended up doing was.
1073
+
1074
+ 00:16:33.470 --> 00:16:37.040
1075
+ Venting Airbnbs in Boston and other
1076
+
1077
+ 00:16:37.040 --> 00:16:39.710
1078
+ places, setting up Alexis, they're
1079
+
1080
+ 00:16:39.710 --> 00:16:42.488
1081
+ creating scripts and hiring people to
1082
+
1083
+ 00:16:42.488 --> 00:16:44.672
1084
+ walk through these Airbnbs reading the
1085
+
1086
+ 00:16:44.672 --> 00:16:46.495
1087
+ scripts, reading a certain script in
1088
+
1089
+ 00:16:46.495 --> 00:16:49.490
1090
+ each room so that way they know what is
1091
+
1092
+ 00:16:49.490 --> 00:16:51.785
1093
+ said in each room, so they have the
1094
+
1095
+ 00:16:51.785 --> 00:16:52.355
1096
+ ground truth.
1097
+
1098
+ 00:16:52.355 --> 00:16:54.252
1099
+ The script is basically the ground
1100
+
1101
+ 00:16:54.252 --> 00:16:56.470
1102
+ truth, and they can automatically train
1103
+
1104
+ 00:16:56.470 --> 00:16:59.542
1105
+ and test their systems on all of these
1106
+
1107
+ 00:16:59.542 --> 00:17:02.160
1108
+ people that they on the voices of all
1109
+
1110
+ 00:17:02.160 --> 00:17:03.446
1111
+ the people they hired to walk through
1112
+
1113
+ 00:17:03.446 --> 00:17:04.160
1114
+ all these different.
1115
+
1116
+ 00:17:04.220 --> 00:17:04.940
1117
+ Apartments.
1118
+
1119
+ 00:17:04.940 --> 00:17:08.370
1120
+ And so they were able to thousand X
1121
+
1122
+ 00:17:08.370 --> 00:17:11.100
1123
+ their training examples and that's how
1124
+
1125
+ 00:17:11.100 --> 00:17:13.670
1126
+ they achieved their robust speech
1127
+
1128
+ 00:17:13.670 --> 00:17:14.820
1129
+ recognition that they have.
1130
+
1131
+ 00:17:16.400 --> 00:17:20.160
1132
+ And so it's kind of the lesson is that
1133
+
1134
+ 00:17:20.160 --> 00:17:22.430
1135
+ machine learning is really, it's not
1136
+
1137
+ 00:17:22.430 --> 00:17:25.062
1138
+ just about developing algorithms or
1139
+
1140
+ 00:17:25.062 --> 00:17:28.299
1141
+ about tuning your losses or tuning your
1142
+
1143
+ 00:17:28.300 --> 00:17:30.690
1144
+ parameters, but it's also about data
1145
+
1146
+ 00:17:30.690 --> 00:17:32.620
1147
+ preparation and the actual deployment
1148
+
1149
+ 00:17:32.620 --> 00:17:34.420
1150
+ of the technology.
1151
+
1152
+ 00:17:37.840 --> 00:17:39.590
1153
+ So this.
1154
+
1155
+ 00:17:40.850 --> 00:17:42.860
1156
+ So again, like in this class, we're
1157
+
1158
+ 00:17:42.860 --> 00:17:44.690
1159
+ going to focus on the Algorithm and
1160
+
1161
+ 00:17:44.690 --> 00:17:46.610
1162
+ model development and partly it's
1163
+
1164
+ 00:17:46.610 --> 00:17:48.370
1165
+ because there's not time to go out and
1166
+
1167
+ 00:17:48.370 --> 00:17:49.855
1168
+ collect lots of data and Annotate it.
1169
+
1170
+ 00:17:49.855 --> 00:17:51.420
1171
+ It's extremely time consuming.
1172
+
1173
+ 00:17:51.420 --> 00:17:53.306
1174
+ So this is what we can focus on and it
1175
+
1176
+ 00:17:53.306 --> 00:17:55.360
1177
+ is the most challenging and technically
1178
+
1179
+ 00:17:55.360 --> 00:17:58.949
1180
+ complex part of the development.
1181
+
1182
+ 00:18:01.360 --> 00:18:03.910
1183
+ A lot of most, if not all, machine
1184
+
1185
+ 00:18:03.910 --> 00:18:06.270
1186
+ learning algorithms fall into this very
1187
+
1188
+ 00:18:06.270 --> 00:18:07.790
1189
+ simple diagram.
1190
+
1191
+ 00:18:08.550 --> 00:18:10.690
1192
+ Where you have some Raw data, it could
1193
+
1194
+ 00:18:10.690 --> 00:18:13.330
1195
+ be text or images or audio or
1196
+
1197
+ 00:18:13.330 --> 00:18:15.500
1198
+ temperatures or other information.
1199
+
1200
+ 00:18:16.520 --> 00:18:19.466
1201
+ Usually that Raw data is not in a form
1202
+
1203
+ 00:18:19.466 --> 00:18:22.210
1204
+ that is very useful for Prediction.
1205
+
1206
+ 00:18:22.210 --> 00:18:24.502
1207
+ For example, if you receive an image,
1208
+
1209
+ 00:18:24.502 --> 00:18:27.010
1210
+ the intensities of the image are just
1211
+
1212
+ 00:18:27.010 --> 00:18:28.780
1213
+ numbers that correspond to how bright
1214
+
1215
+ 00:18:28.780 --> 00:18:31.410
1216
+ each pixel is, and individually those
1217
+
1218
+ 00:18:31.410 --> 00:18:32.860
1219
+ numbers don't really tell you much at
1220
+
1221
+ 00:18:32.860 --> 00:18:33.530
1222
+ all.
1223
+
1224
+ 00:18:33.530 --> 00:18:35.296
1225
+ What's more meaningful are the
1226
+
1227
+ 00:18:35.296 --> 00:18:37.550
1228
+ patterns, the local patterns of the
1229
+
1230
+ 00:18:37.550 --> 00:18:37.950
1231
+ intensities.
1232
+
1233
+ 00:18:38.810 --> 00:18:40.590
1234
+ And so you need to have some kind of
1235
+
1236
+ 00:18:40.590 --> 00:18:44.110
1237
+ encoding of that Raw data that makes
1238
+
1239
+ 00:18:44.110 --> 00:18:45.370
1240
+ the data more meaningful.
1241
+
1242
+ 00:18:46.340 --> 00:18:49.240
1243
+ And there's lots of different ways that
1244
+
1245
+ 00:18:49.240 --> 00:18:50.530
1246
+ you can create that encoding.
1247
+
1248
+ 00:18:50.530 --> 00:18:52.315
1249
+ You can manually define features, you
1250
+
1251
+ 00:18:52.315 --> 00:18:54.870
1252
+ could have Trees, you could do
1253
+
1254
+ 00:18:54.870 --> 00:18:57.060
1255
+ Probabilistic estimation, or use Deep
1256
+
1257
+ 00:18:57.060 --> 00:18:57.850
1258
+ networks.
1259
+
1260
+ 00:18:58.600 --> 00:18:59.715
1261
+ Then you have a Decoder.
1262
+
1263
+ 00:18:59.715 --> 00:19:01.210
1264
+ The Decoder takes your encoded
1265
+
1266
+ 00:19:01.210 --> 00:19:03.480
1267
+ representation and then it tries to
1268
+
1269
+ 00:19:03.480 --> 00:19:05.060
1270
+ produce something useful from it.
1271
+
1272
+ 00:19:05.060 --> 00:19:07.120
1273
+ Some Prediction that you want classify
1274
+
1275
+ 00:19:07.120 --> 00:19:10.910
1276
+ an image or convert the raw audio into
1277
+
1278
+ 00:19:10.910 --> 00:19:11.390
1279
+ text.
1280
+
1281
+ 00:19:12.680 --> 00:19:14.663
1282
+ Again, there's lots of decoders.
1283
+
1284
+ 00:19:14.663 --> 00:19:16.790
1285
+ There's actually not so much complexity
1286
+
1287
+ 00:19:16.790 --> 00:19:18.910
1288
+ usually in the decoders, but some of
1289
+
1290
+ 00:19:18.910 --> 00:19:20.800
1291
+ the common ones are nearest neighbor or
1292
+
1293
+ 00:19:20.800 --> 00:19:24.190
1294
+ Linear decoders, an SVM or logistic,
1295
+
1296
+ 00:19:24.190 --> 00:19:25.220
1297
+ logistic regressor.
1298
+
1299
+ 00:19:26.960 --> 00:19:28.560
1300
+ Then the decoders will produce some
1301
+
1302
+ 00:19:28.560 --> 00:19:30.630
1303
+ Prediction from the encoded Raw data.
1304
+
1305
+ 00:19:30.630 --> 00:19:32.480
1306
+ And again, there's lots of different
1307
+
1308
+ 00:19:32.480 --> 00:19:33.670
1309
+ kinds of problems out there, so there
1310
+
1311
+ 00:19:33.670 --> 00:19:34.850
1312
+ are many different things that you
1313
+
1314
+ 00:19:34.850 --> 00:19:35.820
1315
+ might be trying to predict.
1316
+
1317
+ 00:19:35.820 --> 00:19:37.610
1318
+ Could be binary labels, you could be
1319
+
1320
+ 00:19:37.610 --> 00:19:39.710
1321
+ producing an image, or producing Text,
1322
+
1323
+ 00:19:39.710 --> 00:19:41.690
1324
+ or detecting objects.
1325
+
1326
+ 00:19:42.620 --> 00:19:44.666
1327
+ When you're training, you'll also have
1328
+
1329
+ 00:19:44.666 --> 00:19:46.330
1330
+ some target Labels, so you have
1331
+
1332
+ 00:19:46.330 --> 00:19:47.000
1333
+ annotated data.
1334
+
1335
+ 00:19:47.000 --> 00:19:48.390
1336
+ If you have a Supervised Algorithm
1337
+
1338
+ 00:19:48.390 --> 00:19:51.150
1339
+ anyway, you have some annotated data
1340
+
1341
+ 00:19:51.150 --> 00:19:52.426
1342
+ where you have the Raw data and you
1343
+
1344
+ 00:19:52.426 --> 00:19:53.580
1345
+ have the labels that you want to
1346
+
1347
+ 00:19:53.580 --> 00:19:54.390
1348
+ predict.
1349
+
1350
+ 00:19:54.390 --> 00:19:56.410
1351
+ You see how different your target
1352
+
1353
+ 00:19:56.410 --> 00:19:58.650
1354
+ Labels are from the predictions, and
1355
+
1356
+ 00:19:58.650 --> 00:20:00.390
1357
+ then based on that you.
1358
+
1359
+ 00:20:01.190 --> 00:20:04.225
1360
+ Improve your Decoder, and in some cases
1361
+
1362
+ 00:20:04.225 --> 00:20:06.150
1363
+ you can then feed that back into your
1364
+
1365
+ 00:20:06.150 --> 00:20:08.600
1366
+ Encoder to improve your encoding
1367
+
1368
+ 00:20:08.600 --> 00:20:10.270
1369
+ representation.
1370
+
1371
+ 00:20:10.270 --> 00:20:12.520
1372
+ So there's a single process that is
1373
+
1374
+ 00:20:12.520 --> 00:20:14.890
1375
+ used by almost all of machine learning,
1376
+
1377
+ 00:20:14.890 --> 00:20:15.990
1378
+ but there's a lot of different
1379
+
1380
+ 00:20:15.990 --> 00:20:17.850
1381
+ variation and a lot of different
1382
+
1383
+ 00:20:17.850 --> 00:20:18.600
1384
+ details.
1385
+
1386
+ 00:20:18.600 --> 00:20:20.414
1387
+ And that's just because of all the
1388
+
1389
+ 00:20:20.414 --> 00:20:21.897
1390
+ different kinds of Raw data that you
1391
+
1392
+ 00:20:21.897 --> 00:20:23.864
1393
+ could be dealing with in all the
1394
+
1395
+ 00:20:23.864 --> 00:20:25.213
1396
+ different kinds of predictions that you
1397
+
1398
+ 00:20:25.213 --> 00:20:25.730
1399
+ could do.
1400
+
1401
+ 00:20:29.380 --> 00:20:31.157
1402
+ So that's machine learning.
1403
+
1404
+ 00:20:31.157 --> 00:20:33.140
1405
+ That's machine learning in general.
1406
+
1407
+ 00:20:33.140 --> 00:20:34.470
1408
+ And so now I'm going to talk a little
1409
+
1410
+ 00:20:34.470 --> 00:20:36.020
1411
+ bit about some of the details of the
1412
+
1413
+ 00:20:36.020 --> 00:20:36.730
1414
+ course.
1415
+
1416
+ 00:20:37.780 --> 00:20:40.500
1417
+ So one of the one of the main course
1418
+
1419
+ 00:20:40.500 --> 00:20:42.680
1420
+ objectives is to try to learn how to
1421
+
1422
+ 00:20:42.680 --> 00:20:44.420
1423
+ solve problems with machine learning.
1424
+
1425
+ 00:20:45.110 --> 00:20:47.990
1426
+ So this involves trying to learn the
1427
+
1428
+ 00:20:47.990 --> 00:20:50.132
1429
+ key concepts and methodologies for how
1430
+
1431
+ 00:20:50.132 --> 00:20:51.140
1432
+ we can learn from data.
1433
+
1434
+ 00:20:51.140 --> 00:20:53.170
1435
+ We're going to learn about a wide
1436
+
1437
+ 00:20:53.170 --> 00:20:56.230
1438
+ variety of algorithms and what they're
1439
+
1440
+ 00:20:56.230 --> 00:20:57.890
1441
+ strengths and limitations are.
1442
+
1443
+ 00:20:57.890 --> 00:21:00.110
1444
+ And we're also going to learn about
1445
+
1446
+ 00:21:00.110 --> 00:21:02.138
1447
+ domain specific representations like
1448
+
1449
+ 00:21:02.138 --> 00:21:04.090
1450
+ how we can encode images, how we can
1451
+
1452
+ 00:21:04.090 --> 00:21:06.220
1453
+ encode text, how we can encode audio.
1454
+
1455
+ 00:21:06.220 --> 00:21:10.095
1456
+ And then we want to at the end of this
1457
+
1458
+ 00:21:10.095 --> 00:21:11.540
1459
+ have the ability to select the right
1460
+
1461
+ 00:21:11.540 --> 00:21:13.870
1462
+ tools for the job so that if you have
1463
+
1464
+ 00:21:13.870 --> 00:21:15.530
1465
+ your own custom problem that.
1466
+
1467
+ 00:21:15.580 --> 00:21:17.180
1468
+ You want to solve that.
1469
+
1470
+ 00:21:17.180 --> 00:21:19.270
1471
+ You're able to make good choices about
1472
+
1473
+ 00:21:19.270 --> 00:21:22.000
1474
+ how to represent the data and what
1475
+
1476
+ 00:21:22.000 --> 00:21:24.160
1477
+ kinds of algorithms and losses to use,
1478
+
1479
+ 00:21:24.160 --> 00:21:26.812
1480
+ how to optimize your algorithms, and
1481
+
1482
+ 00:21:26.812 --> 00:21:29.140
1483
+ how to solve your problem.
1484
+
1485
+ 00:21:31.010 --> 00:21:34.049
1486
+ So I think this is, I think it's just
1487
+
1488
+ 00:21:34.050 --> 00:21:36.510
1489
+ generally interesting to be able to do
1490
+
1491
+ 00:21:36.510 --> 00:21:38.450
1492
+ this, but it's also very practical.
1493
+
1494
+ 00:21:38.450 --> 00:21:41.280
1495
+ One example is just if you look at it
1496
+
1497
+ 00:21:41.280 --> 00:21:43.880
1498
+ in terms of the number of jobs that are
1499
+
1500
+ 00:21:43.880 --> 00:21:45.930
1501
+ available in the demand for people that
1502
+
1503
+ 00:21:45.930 --> 00:21:48.250
1504
+ have skills in machine learning
1505
+
1506
+ 00:21:48.250 --> 00:21:50.810
1507
+ engineering, it's an area that's been
1508
+
1509
+ 00:21:50.810 --> 00:21:52.260
1510
+ growing really rapidly.
1511
+
1512
+ 00:21:52.260 --> 00:21:55.469
1513
+ So according to this, according to this
1514
+
1515
+ 00:21:55.470 --> 00:21:55.910
1516
+ source.
1517
+
1518
+ 00:21:56.540 --> 00:21:58.220
1519
+ The.
1520
+
1521
+ 00:21:58.310 --> 00:22:00.380
1522
+ Market for global machine learning is
1523
+
1524
+ 00:22:00.380 --> 00:22:02.860
1525
+ expected to grow from around 20 billion
1526
+
1527
+ 00:22:02.860 --> 00:22:07.230
1528
+ last year to 200 billion in 2029, which
1529
+
1530
+ 00:22:07.230 --> 00:22:09.080
1531
+ is a growth rate of about 38%.
1532
+
1533
+ 00:22:09.080 --> 00:22:11.340
1534
+ So machine learning engineers are in
1535
+
1536
+ 00:22:11.340 --> 00:22:13.300
1537
+ high demand now and they're going to
1538
+
1539
+ 00:22:13.300 --> 00:22:14.959
1540
+ likely continue to be in high demand
1541
+
1542
+ 00:22:14.960 --> 00:22:16.340
1543
+ over the next several years.
1544
+
1545
+ 00:22:18.860 --> 00:22:21.540
1546
+ Another objective of the course is to
1547
+
1548
+ 00:22:21.540 --> 00:22:23.010
1549
+ have a better understanding of the real
1550
+
1551
+ 00:22:23.010 --> 00:22:25.150
1552
+ life applications and the social
1553
+
1554
+ 00:22:25.150 --> 00:22:26.758
1555
+ implications of machine learning.
1556
+
1557
+ 00:22:26.758 --> 00:22:29.330
1558
+ So I'll try to talk about actual
1559
+
1560
+ 00:22:29.330 --> 00:22:30.720
1561
+ deployments and machine learning and
1562
+
1563
+ 00:22:30.720 --> 00:22:31.712
1564
+ several cases.
1565
+
1566
+ 00:22:31.712 --> 00:22:35.870
1567
+ We'll talk about some of the ethical
1568
+
1569
+ 00:22:35.870 --> 00:22:37.479
1570
+ concerns around machine learning and
1571
+
1572
+ 00:22:37.480 --> 00:22:40.899
1573
+ its use, and about the and about
1574
+
1575
+ 00:22:40.900 --> 00:22:42.390
1576
+ different application domains.
1577
+
1578
+ 00:22:42.390 --> 00:22:44.440
1579
+ So machine learning can do a lot of
1580
+
1581
+ 00:22:44.440 --> 00:22:45.050
1582
+ good.
1583
+
1584
+ 00:22:45.050 --> 00:22:46.800
1585
+ There's a lot of ways that.
1586
+
1587
+ 00:22:46.860 --> 00:22:49.170
1588
+ Already impacts our lives through our
1589
+
1590
+ 00:22:49.170 --> 00:22:52.330
1591
+ cameras, through through Smart
1592
+
1593
+ 00:22:52.330 --> 00:22:54.283
1594
+ assistants and things like that.
1595
+
1596
+ 00:22:54.283 --> 00:22:58.193
1597
+ Of course, technology can also create a
1598
+
1599
+ 00:22:58.193 --> 00:23:00.453
1600
+ lot of upheaval and also can do a lot
1601
+
1602
+ 00:23:00.453 --> 00:23:01.230
1603
+ of damage.
1604
+
1605
+ 00:23:01.230 --> 00:23:03.820
1606
+ And so it's important to understand the
1607
+
1608
+ 00:23:03.820 --> 00:23:06.390
1609
+ implications of the technology that we
1610
+
1611
+ 00:23:06.390 --> 00:23:06.940
1612
+ develop.
1613
+
1614
+ 00:23:09.670 --> 00:23:11.630
1615
+ And then the third is appreciation for
1616
+
1617
+ 00:23:11.630 --> 00:23:13.160
1618
+ your own constantly Learning minds.
1619
+
1620
+ 00:23:13.160 --> 00:23:15.480
1621
+ So we're humans are still the best
1622
+
1623
+ 00:23:15.480 --> 00:23:17.180
1624
+ machine Learners, I would say.
1625
+
1626
+ 00:23:18.290 --> 00:23:20.490
1627
+ You're always you're always learning
1628
+
1629
+ 00:23:20.490 --> 00:23:21.370
1630
+ new things.
1631
+
1632
+ 00:23:21.370 --> 00:23:23.170
1633
+ You're always learning about new
1634
+
1635
+ 00:23:23.170 --> 00:23:24.140
1636
+ environments.
1637
+
1638
+ 00:23:24.140 --> 00:23:27.730
1639
+ You learn new skills, new facts, new
1640
+
1641
+ 00:23:27.730 --> 00:23:28.720
1642
+ concepts.
1643
+
1644
+ 00:23:28.720 --> 00:23:34.666
1645
+ And the flexibility of our Learning and
1646
+
1647
+ 00:23:34.666 --> 00:23:37.140
1648
+ the breadth of it and the seamlessness
1649
+
1650
+ 00:23:37.140 --> 00:23:39.040
1651
+ of it is really amazing.
1652
+
1653
+ 00:23:39.040 --> 00:23:42.420
1654
+ And so as you learn how, even though we
1655
+
1656
+ 00:23:42.420 --> 00:23:46.290
1657
+ can do pretty cool things with machine
1658
+
1659
+ 00:23:46.290 --> 00:23:47.940
1660
+ learning and computer science.
1661
+
1662
+ 00:23:48.380 --> 00:23:51.140
1663
+ It's much more so far.
1664
+
1665
+ 00:23:51.140 --> 00:23:52.880
1666
+ Machine learning is much more rigid and
1667
+
1668
+ 00:23:52.880 --> 00:23:55.370
1669
+ focused and.
1670
+
1671
+ 00:23:55.520 --> 00:23:56.850
1672
+ And challenging.
1673
+
1674
+ 00:23:56.850 --> 00:24:00.675
1675
+ And so you can think about how did you
1676
+
1677
+ 00:24:00.675 --> 00:24:02.520
1678
+ learn to do things, how do you learn to
1679
+
1680
+ 00:24:02.520 --> 00:24:03.144
1681
+ make predictions?
1682
+
1683
+ 00:24:03.144 --> 00:24:05.760
1684
+ And think about how it may be similar
1685
+
1686
+ 00:24:05.760 --> 00:24:07.940
1687
+ or different from the machine learning
1688
+
1689
+ 00:24:07.940 --> 00:24:09.480
1690
+ algorithms that we learn about.
1691
+
1692
+ 00:24:12.410 --> 00:24:13.760
1693
+ So.
1694
+
1695
+ 00:24:14.230 --> 00:24:15.800
1696
+ All right, so now I'm getting more into
1697
+
1698
+ 00:24:15.800 --> 00:24:18.030
1699
+ the logistics of the course I
1700
+
1701
+ 00:24:18.030 --> 00:24:19.280
1702
+ introduced myself already.
1703
+
1704
+ 00:24:19.280 --> 00:24:21.100
1705
+ I want to introduce a few of the tasks
1706
+
1707
+ 00:24:21.100 --> 00:24:21.660
1708
+ that are here.
1709
+
1710
+ 00:24:21.660 --> 00:24:23.380
1711
+ There's one traveling and one sick.
1712
+
1713
+ 00:24:24.700 --> 00:24:26.960
1714
+ But some of them are here, so one is
1715
+
1716
+ 00:24:26.960 --> 00:24:29.120
1717
+ Vatsal Chheda.
1718
+
1719
+ 00:24:29.120 --> 00:24:30.650
1720
+ Could you introduce yourself?
1721
+
1722
+ 00:24:32.750 --> 00:24:34.880
1723
+ And I think I should probably give you
1724
+
1725
+ 00:24:34.880 --> 00:24:35.570
1726
+ the mic.
1727
+
1728
+ 00:24:39.030 --> 00:24:40.740
1729
+ If I can, sort of.
1730
+
1731
+ 00:24:46.550 --> 00:24:49.735
1732
+ Hello everyone my name is Vatsal, I'm
1733
+
1734
+ 00:24:49.735 --> 00:24:53.290
1735
+ from India and the like the
1736
+
1737
+ 00:24:53.290 --> 00:24:54.490
1738
+ applications of applied machine
1739
+
1740
+ 00:24:54.490 --> 00:24:55.070
1741
+ learning are.
1742
+
1743
+ 00:24:55.070 --> 00:24:56.390
1744
+ I'm using it in neural machine
1745
+
1746
+ 00:24:56.390 --> 00:24:58.460
1747
+ translation from Polish language to
1748
+
1749
+ 00:24:58.460 --> 00:25:00.630
1750
+ English language and in various NLP
1751
+
1752
+ 00:25:00.630 --> 00:25:01.420
1753
+ projects.
1754
+
1755
+ 00:25:02.740 --> 00:25:03.540
1756
+ Thank you.
1757
+
1758
+ 00:25:05.100 --> 00:25:08.220
1759
+ I think Josh is out sick today, Weijie.
1760
+
1761
+ 00:25:14.790 --> 00:25:16.900
1762
+ Hi everyone, I'm Weijie.
1763
+
1764
+ 00:25:16.900 --> 00:25:19.235
1765
+ I'm a second year Masters student and
1766
+
1767
+ 00:25:19.235 --> 00:25:20.565
1768
+ I'm from China.
1769
+
1770
+ 00:25:20.565 --> 00:25:23.920
1771
+ So I basically do research about
1772
+
1773
+ 00:25:23.920 --> 00:25:26.860
1774
+ computer vision and machine learning.
1775
+
1776
+ 00:25:26.860 --> 00:25:28.510
1777
+ So nice to meet you all.
1778
+
1779
+ 00:25:28.510 --> 00:25:29.120
1780
+ Thank you.
1781
+
1782
+ 00:25:31.120 --> 00:25:32.440
1783
+ OK.
1784
+
1785
+ 00:25:32.440 --> 00:25:34.720
1786
+ And since each units here, I'll go to
1787
+
1788
+ 00:25:34.720 --> 00:25:35.570
1789
+ YouTube next.
1790
+
1791
+ 00:25:39.540 --> 00:25:41.770
1792
+ Hi everyone I'm reaching you and I'm a
1793
+
1794
+ 00:25:41.770 --> 00:25:44.863
1795
+ first Masters student in the group and
1796
+
1797
+ 00:25:44.863 --> 00:25:48.420
1798
+ I came from Shenzhen City located in
1799
+
1800
+ 00:25:48.420 --> 00:25:51.630
1801
+ southern China and I'm interested in 3D
1802
+
1803
+ 00:25:51.630 --> 00:25:53.240
1804
+ reconstruction and 3D scene
1805
+
1806
+ 00:25:53.240 --> 00:25:53.850
1807
+ understanding.
1808
+
1809
+ 00:25:53.850 --> 00:25:56.160
1810
+ And during my research I have utilized
1811
+
1812
+ 00:25:56.160 --> 00:25:59.400
1813
+ some deep neural network, RESNET and
1814
+
1815
+ 00:25:59.400 --> 00:26:01.040
1816
+ multi layer perception and they are
1817
+
1818
+ 00:26:01.040 --> 00:26:04.010
1819
+ really useful and can produce pretty
1820
+
1821
+ 00:26:04.010 --> 00:26:04.910
1822
+ decent results.
1823
+
1824
+ 00:26:04.910 --> 00:26:06.550
1825
+ Looking forward to learning with you.
1826
+
1827
+ 00:26:06.610 --> 00:26:07.150
1828
+ Thank you.
1829
+
1830
+ 00:26:10.220 --> 00:26:12.630
1831
+ All right and.
1832
+
1833
+ 00:26:12.990 --> 00:26:16.420
1834
+ OK, **** is traveling and let's see
1835
+
1836
+ 00:26:16.420 --> 00:26:17.080
1837
+ Mington.
1838
+
1839
+ 00:26:21.990 --> 00:26:24.343
1840
+ Hi everyone, my name is Min Hongchang
1841
+
1842
+ 00:26:24.343 --> 00:26:26.820
1843
+ and I'm a first year masters master
1844
+
1845
+ 00:26:26.820 --> 00:26:28.790
1846
+ student in computer science and I'm
1847
+
1848
+ 00:26:28.790 --> 00:26:33.210
1849
+ from China and my current research is
1850
+
1851
+ 00:26:33.210 --> 00:26:35.380
1852
+ focused on computer vision, especially
1853
+
1854
+ 00:26:35.380 --> 00:26:38.515
1855
+ for 3D Vision and generative modeling.
1856
+
1857
+ 00:26:38.515 --> 00:26:41.690
1858
+ And I hope you all enjoyed this course.
1859
+
1860
+ 00:26:41.690 --> 00:26:42.380
1861
+ Thank you.
1862
+
1863
+ 00:26:43.010 --> 00:26:45.090
1864
+ Here and Wentao.
1865
+
1866
+ 00:26:51.480 --> 00:26:54.630
1867
+ So on Wentao and I'm a second year
1868
+
1869
+ 00:26:54.630 --> 00:26:57.080
1870
+ master student here and last semester
1871
+
1872
+ 00:26:57.080 --> 00:27:01.095
1873
+ also tax CS440 if someone take it.
1874
+
1875
+ 00:27:01.095 --> 00:27:04.510
1876
+ And my research mainly focus on any
1877
+
1878
+ 00:27:04.510 --> 00:27:06.380
1879
+ kind of sequential Prediction like
1880
+
1881
+ 00:27:06.380 --> 00:27:07.560
1882
+ natural language understanding,
1883
+
1884
+ 00:27:07.560 --> 00:27:09.840
1885
+ national generation, even some
1886
+
1887
+ 00:27:09.840 --> 00:27:12.670
1888
+ sequential Prediction in India Vision
1889
+
1890
+ 00:27:12.670 --> 00:27:15.070
1891
+ post like trajectory.
1892
+
1893
+ 00:27:15.130 --> 00:27:17.540
1894
+ So if you have any question about that
1895
+
1896
+ 00:27:17.540 --> 00:27:18.600
1897
+ you can ask me.
1898
+
1899
+ 00:27:21.050 --> 00:27:22.085
1900
+ Alright, thank you.
1901
+
1902
+ 00:27:22.085 --> 00:27:25.510
1903
+ And I'll have Josh and kitchenette
1904
+
1905
+ 00:27:25.510 --> 00:27:27.090
1906
+ introduced themselves.
1907
+
1908
+ 00:27:28.350 --> 00:27:29.690
1909
+ When they another time.
1910
+
1911
+ 00:27:34.360 --> 00:27:34.830
1912
+ All right.
1913
+
1914
+ 00:27:34.830 --> 00:27:36.490
1915
+ So there's.
1916
+
1917
+ 00:27:37.730 --> 00:27:39.526
1918
+ So there's three main topics that we're
1919
+
1920
+ 00:27:39.526 --> 00:27:40.080
1921
+ going to cover.
1922
+
1923
+ 00:27:40.080 --> 00:27:43.090
1924
+ One is Supervised Learning
1925
+
1926
+ 00:27:43.090 --> 00:27:43.850
1927
+ Fundamentals.
1928
+
1929
+ 00:27:43.850 --> 00:27:46.346
1930
+ So we'll do the next lecture is going
1931
+
1932
+ 00:27:46.346 --> 00:27:48.890
1933
+ to be on KNN's and then we'll talk
1934
+
1935
+ 00:27:48.890 --> 00:27:52.160
1936
+ about Naive Bayes algorithm and be kind
1937
+
1938
+ 00:27:52.160 --> 00:27:55.406
1939
+ of a review of probability as well.
1940
+
1941
+ 00:27:55.406 --> 00:27:58.170
1942
+ Then linear and logistic regression
1943
+
1944
+ 00:27:58.170 --> 00:28:01.180
1945
+ trees and random forests, and maybe
1946
+
1947
+ 00:28:01.180 --> 00:28:05.383
1948
+ boosting ensemble methods, SVMs, neural
1949
+
1950
+ 00:28:05.383 --> 00:28:07.070
1951
+ networks and deep neural networks.
1952
+
1953
+ 00:28:07.900 --> 00:28:09.590
1954
+ Then the next section, the class, we're
1955
+
1956
+ 00:28:09.590 --> 00:28:10.880
1957
+ going to talk about some different
1958
+
1959
+ 00:28:10.880 --> 00:28:12.010
1960
+ application domains.
1961
+
1962
+ 00:28:12.910 --> 00:28:15.447
1963
+ So we'll talk about computer vision and
1964
+
1965
+ 00:28:15.447 --> 00:28:19.320
1966
+ using CNS for computer vision language
1967
+
1968
+ 00:28:19.320 --> 00:28:21.770
1969
+ Models, will talk about using the
1970
+
1971
+ 00:28:21.770 --> 00:28:24.424
1972
+ transformer networks for vision and
1973
+
1974
+ 00:28:24.424 --> 00:28:24.930
1975
+ language.
1976
+
1977
+ 00:28:24.930 --> 00:28:26.720
1978
+ This idea that you may have heard about
1979
+
1980
+ 00:28:26.720 --> 00:28:28.200
1981
+ called Foundation Models where you
1982
+
1983
+ 00:28:28.200 --> 00:28:30.789
1984
+ where you create machine, where you
1985
+
1986
+ 00:28:30.790 --> 00:28:32.280
1987
+ train machine learning models on a lot
1988
+
1989
+ 00:28:32.280 --> 00:28:33.855
1990
+ of data and then you kind of adapt it
1991
+
1992
+ 00:28:33.855 --> 00:28:34.565
1993
+ for other tasks.
1994
+
1995
+ 00:28:34.565 --> 00:28:37.070
1996
+ And so we'll talk about task and the
1997
+
1998
+ 00:28:37.070 --> 00:28:41.800
1999
+ main adaptation audio as well the
2000
+
2001
+ 00:28:41.800 --> 00:28:42.500
2002
+ ethics.
2003
+
2004
+ 00:28:42.570 --> 00:28:43.120
2005
+ Issues.
2006
+
2007
+ 00:28:43.120 --> 00:28:45.600
2008
+ Really ethical issues relating to
2009
+
2010
+ 00:28:45.600 --> 00:28:46.310
2011
+ machine learning.
2012
+
2013
+ 00:28:47.040 --> 00:28:48.540
2014
+ And data.
2015
+
2016
+ 00:28:49.460 --> 00:28:50.620
2017
+ And then the final section.
2018
+
2019
+ 00:28:50.620 --> 00:28:53.150
2020
+ The class is on parent Discovery and
2021
+
2022
+ 00:28:53.150 --> 00:28:54.990
2023
+ that focuses more on the unsupervised
2024
+
2025
+ 00:28:54.990 --> 00:28:56.990
2026
+ method, so methods for Clustering and
2027
+
2028
+ 00:28:56.990 --> 00:28:59.290
2029
+ retrieval, missing data and the
2030
+
2031
+ 00:28:59.290 --> 00:29:01.290
2032
+ expectation maximization algorithm
2033
+
2034
+ 00:29:01.290 --> 00:29:02.530
2035
+ Density estimation.
2036
+
2037
+ 00:29:03.590 --> 00:29:06.090
2038
+ Creating Topic Models, dealing with
2039
+
2040
+ 00:29:06.090 --> 00:29:08.710
2041
+ outliers and data visualization.
2042
+
2043
+ 00:29:08.710 --> 00:29:11.390
2044
+ CCA, So some of the details might
2045
+
2046
+ 00:29:11.390 --> 00:29:13.840
2047
+ change, especially towards the end of
2048
+
2049
+ 00:29:13.840 --> 00:29:14.360
2050
+ the class.
2051
+
2052
+ 00:29:14.360 --> 00:29:16.960
2053
+ I'm not setting it in stone yet because
2054
+
2055
+ 00:29:16.960 --> 00:29:20.110
2056
+ I'll see how things evolve and see if I
2057
+
2058
+ 00:29:20.110 --> 00:29:22.690
2059
+ have better ideas later, but at least
2060
+
2061
+ 00:29:22.690 --> 00:29:26.780
2062
+ the initial few weeks are more or less
2063
+
2064
+ 00:29:26.780 --> 00:29:29.090
2065
+ determined in the schedules online, as
2066
+
2067
+ 00:29:29.090 --> 00:29:29.800
2068
+ I'll show.
2069
+
2070
+ 00:29:32.380 --> 00:29:33.170
2071
+ So.
2072
+
2073
+ 00:29:33.260 --> 00:29:33.920
2074
+
2075
+
2076
+ 00:29:37.340 --> 00:29:37.600
2077
+ Don't.
2078
+
2079
+ 00:29:37.600 --> 00:29:39.410
2080
+ I don't usually have people cheer when
2081
+
2082
+ 00:29:39.410 --> 00:29:41.950
2083
+ I show the show the Assignments.
2084
+
2085
+ 00:29:43.180 --> 00:29:45.690
2086
+ Alright, so there's 222 different
2087
+
2088
+ 00:29:45.690 --> 00:29:46.850
2089
+ components to your grades.
2090
+
2091
+ 00:29:46.850 --> 00:29:48.670
2092
+ There's homeworks and Final projects.
2093
+
2094
+ 00:29:48.670 --> 00:29:50.700
2095
+ There's four signed homeworks.
2096
+
2097
+ 00:29:50.700 --> 00:29:52.640
2098
+ Those are worth at least 100 points
2099
+
2100
+ 00:29:52.640 --> 00:29:53.000
2101
+ each.
2102
+
2103
+ 00:29:53.710 --> 00:29:56.190
2104
+ So you can see homework one now that's
2105
+
2106
+ 00:29:56.190 --> 00:29:59.930
2107
+ worth 160 points, but the only about
2108
+
2109
+ 00:29:59.930 --> 00:30:01.413
2110
+ 100 is really expected.
2111
+
2112
+ 00:30:01.413 --> 00:30:03.830
2113
+ So for most of the homeworks, there's a
2114
+
2115
+ 00:30:03.830 --> 00:30:07.762
2116
+ core of like 100 points that is that is
2117
+
2118
+ 00:30:07.762 --> 00:30:10.910
2119
+ more prescripted and then there's like
2120
+
2121
+ 00:30:10.910 --> 00:30:13.770
2122
+ additional points for that require more
2123
+
2124
+ 00:30:13.770 --> 00:30:15.110
2125
+ independent exploration.
2126
+
2127
+ 00:30:16.320 --> 00:30:18.550
2128
+ There's a Final project that's worth
2129
+
2130
+ 00:30:18.550 --> 00:30:21.940
2131
+ 100 points, and what I'm planning to do
2132
+
2133
+ 00:30:21.940 --> 00:30:26.270
2134
+ for that is to select a few.
2135
+
2136
+ 00:30:26.380 --> 00:30:29.820
2137
+ A few key challenges, for example,
2138
+
2139
+ 00:30:29.820 --> 00:30:32.335
2140
+ something like the Netflix challenge,
2141
+
2142
+ 00:30:32.335 --> 00:30:34.820
2143
+ and I'll solicit your input on that.
2144
+
2145
+ 00:30:34.820 --> 00:30:36.910
2146
+ And so everyone would have the option
2147
+
2148
+ 00:30:36.910 --> 00:30:38.540
2149
+ of doing one of those challenges.
2150
+
2151
+ 00:30:38.540 --> 00:30:40.660
2152
+ Or you can just do your own custom
2153
+
2154
+ 00:30:40.660 --> 00:30:41.180
2155
+ project.
2156
+
2157
+ 00:30:43.190 --> 00:30:45.150
2158
+ So you can kind of.
2159
+
2160
+ 00:30:47.170 --> 00:30:47.610
2161
+ There is.
2162
+
2163
+ 00:30:47.610 --> 00:30:49.055
2164
+ There's a lot of different students in
2165
+
2166
+ 00:30:49.055 --> 00:30:50.380
2167
+ the class from different backgrounds
2168
+
2169
+ 00:30:50.380 --> 00:30:51.600
2170
+ and with different interests.
2171
+
2172
+ 00:30:51.600 --> 00:30:53.870
2173
+ And so rather than trying to tell
2174
+
2175
+ 00:30:53.870 --> 00:30:56.200
2176
+ everybody, make everybody do exactly
2177
+
2178
+ 00:30:56.200 --> 00:30:57.430
2179
+ the same thing.
2180
+
2181
+ 00:30:57.430 --> 00:31:00.129
2182
+ I generally like to have kind of like
2183
+
2184
+ 00:31:00.130 --> 00:31:02.240
2185
+ structure, but options so that if
2186
+
2187
+ 00:31:02.240 --> 00:31:04.430
2188
+ you're not interested in something or
2189
+
2190
+ 00:31:04.430 --> 00:31:07.449
2191
+ if you're Schedule is really bad for a
2192
+
2193
+ 00:31:07.450 --> 00:31:09.505
2194
+ couple weeks, you can.
2195
+
2196
+ 00:31:09.505 --> 00:31:11.465
2197
+ There's flexibility built in so that
2198
+
2199
+ 00:31:11.465 --> 00:31:12.869
2200
+ you can do the things that make the
2201
+
2202
+ 00:31:12.870 --> 00:31:15.220
2203
+ most sense for you so.
2204
+
2205
+ 00:31:15.380 --> 00:31:16.706
2206
+ If you're in the three credit version
2207
+
2208
+ 00:31:16.706 --> 00:31:18.779
2209
+ of the course, you're graded out of a
2210
+
2211
+ 00:31:18.780 --> 00:31:21.910
2212
+ total of 450 points, which means that
2213
+
2214
+ 00:31:21.910 --> 00:31:24.940
2215
+ you need to do 4 1/2 of those things.
2216
+
2217
+ 00:31:24.940 --> 00:31:25.816
2218
+ But you could.
2219
+
2220
+ 00:31:25.816 --> 00:31:28.835
2221
+ You could, for example, do like 3
2222
+
2223
+ 00:31:28.835 --> 00:31:32.095
2224
+ homeworks and do all the extra points
2225
+
2226
+ 00:31:32.095 --> 00:31:33.440
2227
+ that are available, and then you'd
2228
+
2229
+ 00:31:33.440 --> 00:31:35.130
2230
+ probably be done for the semester.
2231
+
2232
+ 00:31:35.130 --> 00:31:38.386
2233
+ Or you could do like the bare minimum
2234
+
2235
+ 00:31:38.386 --> 00:31:43.479
2236
+ on 3 homeworks, do half and half of
2237
+
2238
+ 00:31:43.480 --> 00:31:45.530
2239
+ another homework in the Final project.
2240
+
2241
+ 00:31:45.590 --> 00:31:47.370
2242
+ Or any combination, it's up to you.
2243
+
2244
+ 00:31:48.540 --> 00:31:50.500
2245
+ And for the four credit version, you
2246
+
2247
+ 00:31:50.500 --> 00:31:54.682
2248
+ have to do like everything and a little
2249
+
2250
+ 00:31:54.682 --> 00:31:55.597
2251
+ bit more.
2252
+
2253
+ 00:31:55.597 --> 00:31:59.670
2254
+ So you have to do the four Assignments
2255
+
2256
+ 00:31:59.670 --> 00:32:02.967
2257
+ a little bit of the extra, some of the,
2258
+
2259
+ 00:32:02.967 --> 00:32:06.730
2260
+ some of the stretch goals and the Final
2261
+
2262
+ 00:32:06.730 --> 00:32:07.280
2263
+ project.
2264
+
2265
+ 00:32:07.280 --> 00:32:09.020
2266
+ But again, there's enough opportunity
2267
+
2268
+ 00:32:09.020 --> 00:32:10.400
2269
+ that you could just do the four
2270
+
2271
+ 00:32:10.400 --> 00:32:12.230
2272
+ homeworks and the stretch goals and be
2273
+
2274
+ 00:32:12.230 --> 00:32:16.073
2275
+ done or do the OR do a more basic job.
2276
+
2277
+ 00:32:16.073 --> 00:32:18.089
2278
+ There's a little bit of.
2279
+
2280
+ 00:32:18.150 --> 00:32:20.730
2281
+ Extra credit available so if you do an
2282
+
2283
+ 00:32:20.730 --> 00:32:21.510
2284
+ extra.
2285
+
2286
+ 00:32:22.400 --> 00:32:24.470
2287
+ You can earn an additional 15 points
2288
+
2289
+ 00:32:24.470 --> 00:32:27.080
2290
+ beyond what's beyond the amount needed
2291
+
2292
+ 00:32:27.080 --> 00:32:29.150
2293
+ for 100% of the homework grade.
2294
+
2295
+ 00:32:29.890 --> 00:32:32.180
2296
+ So I have in the Syllabus, I have like
2297
+
2298
+ 00:32:32.180 --> 00:32:33.870
2299
+ the actual equation for computing your
2300
+
2301
+ 00:32:33.870 --> 00:32:34.120
2302
+ grades.
2303
+
2304
+ 00:32:34.120 --> 00:32:36.310
2305
+ So if there's any confusion about how
2306
+
2307
+ 00:32:36.310 --> 00:32:38.127
2308
+ that works, you can look at that
2309
+
2310
+ 00:32:38.127 --> 00:32:39.486
2311
+ equation and it's pretty.
2312
+
2313
+ 00:32:39.486 --> 00:32:41.130
2314
+ I think it's pretty straightforward, at
2315
+
2316
+ 00:32:41.130 --> 00:32:41.970
2317
+ least once you see that.
2318
+
2319
+ 00:32:43.280 --> 00:32:47.032
2320
+ Late policy likewise, I like to be kind
2321
+
2322
+ 00:32:47.032 --> 00:32:51.537
2323
+ of flexible and mainly the main reason
2324
+
2325
+ 00:32:51.537 --> 00:32:53.620
2326
+ to have late penalties is that I want
2327
+
2328
+ 00:32:53.620 --> 00:32:56.560
2329
+ to keep everybody on track and also for
2330
+
2331
+ 00:32:56.560 --> 00:32:59.950
2332
+ course logistics have things so that we
2333
+
2334
+ 00:32:59.950 --> 00:33:01.359
2335
+ can like kind of grade things and be
2336
+
2337
+ 00:33:01.360 --> 00:33:02.410
2338
+ done with them at some point.
2339
+
2340
+ 00:33:03.770 --> 00:33:06.010
2341
+ So the late policy is that you get up
2342
+
2343
+ 00:33:06.010 --> 00:33:07.730
2344
+ to 10 free late days that you can use
2345
+
2346
+ 00:33:07.730 --> 00:33:09.030
2347
+ on any of these Assignments.
2348
+
2349
+ 00:33:09.790 --> 00:33:12.250
2350
+ The exception may be the Final project,
2351
+
2352
+ 00:33:12.250 --> 00:33:13.930
2353
+ since that's due towards at the end of
2354
+
2355
+ 00:33:13.930 --> 00:33:15.330
2356
+ this semester question.
2357
+
2358
+ 00:33:15.430 --> 00:33:16.840
2359
+ Cortana Final project.
2360
+
2361
+ 00:33:16.840 --> 00:33:19.290
2362
+ Our TV is going to sign us like massive
2363
+
2364
+ 00:33:19.290 --> 00:33:20.070
2365
+ kit like.
2366
+
2367
+ 00:33:27.580 --> 00:33:30.310
2368
+ So the question is, for the Final
2369
+
2370
+ 00:33:30.310 --> 00:33:32.600
2371
+ project, will the TAs create a massive
2372
+
2373
+ 00:33:32.600 --> 00:33:33.690
2374
+ GitHub repository?
2375
+
2376
+ 00:33:34.400 --> 00:33:36.380
2377
+ And share it?
2378
+
2379
+ 00:33:36.380 --> 00:33:37.710
2380
+ Or do you create your own?
2381
+
2382
+ 00:33:38.940 --> 00:33:39.730
2383
+ So.
2384
+
2385
+ 00:33:40.770 --> 00:33:43.070
2386
+ Most likely the TAs will not create
2387
+
2388
+ 00:33:43.070 --> 00:33:44.950
2389
+ anything for the Final project.
2390
+
2391
+ 00:33:44.950 --> 00:33:47.520
2392
+ So you would because there many people
2393
+
2394
+ 00:33:47.520 --> 00:33:49.920
2395
+ may do their custom projects and then
2396
+
2397
+ 00:33:49.920 --> 00:33:52.386
2398
+ you're doing you're working with it
2399
+
2400
+ 00:33:52.386 --> 00:33:54.010
2401
+ could be online, you could start with
2402
+
2403
+ 00:33:54.010 --> 00:33:56.260
2404
+ an online repository or do it from
2405
+
2406
+ 00:33:56.260 --> 00:33:58.150
2407
+ scratch and likewise for the
2408
+
2409
+ 00:33:58.150 --> 00:33:59.040
2410
+ challenges.
2411
+
2412
+ 00:33:59.120 --> 00:34:02.660
2413
+ And if we pick Kaggle challenges for
2414
+
2415
+ 00:34:02.660 --> 00:34:04.440
2416
+ example, there's some, there's a little
2417
+
2418
+ 00:34:04.440 --> 00:34:06.000
2419
+ bit of like infrastructure around
2420
+
2421
+ 00:34:06.000 --> 00:34:08.240
2422
+ those, but mostly for the final
2423
+
2424
+ 00:34:08.240 --> 00:34:10.050
2425
+ projects, I want it to be more
2426
+
2427
+ 00:34:10.050 --> 00:34:10.800
2428
+ open-ended.
2429
+
2430
+ 00:34:10.800 --> 00:34:12.100
2431
+ So yeah.
2432
+
2433
+ 00:34:13.960 --> 00:34:15.690
2434
+ And how did this?
2435
+
2436
+ 00:34:15.690 --> 00:34:17.070
2437
+ Yeah.
2438
+
2439
+ 00:34:19.100 --> 00:34:21.351
2440
+ So the Final project can be done in a
2441
+
2442
+ 00:34:21.351 --> 00:34:23.037
2443
+ group or it can be if you so.
2444
+
2445
+ 00:34:23.037 --> 00:34:25.310
2446
+ If you do a custom project it has I
2447
+
2448
+ 00:34:25.310 --> 00:34:27.145
2449
+ want it has to be in a group.
2450
+
2451
+ 00:34:27.145 --> 00:34:29.060
2452
+ If you do one of the challenges you can
2453
+
2454
+ 00:34:29.060 --> 00:34:30.482
2455
+ either do it individually or in a
2456
+
2457
+ 00:34:30.482 --> 00:34:30.640
2458
+ group.
2459
+
2460
+ 00:34:32.250 --> 00:34:33.380
2461
+ Up to four.
2462
+
2463
+ 00:34:34.740 --> 00:34:35.380
2464
+ Yeah.
2465
+
2466
+ 00:34:37.680 --> 00:34:40.490
2467
+ OK, so back to the late policy.
2468
+
2469
+ 00:34:40.490 --> 00:34:41.690
2470
+ So the.
2471
+
2472
+ 00:34:42.110 --> 00:34:43.835
2473
+ So you get up to 10 free late days.
2474
+
2475
+ 00:34:43.835 --> 00:34:45.460
2476
+ You can use them on any combination of
2477
+
2478
+ 00:34:45.460 --> 00:34:45.678
2479
+ projects.
2480
+
2481
+ 00:34:45.678 --> 00:34:47.400
2482
+ You can be 10 days late for one
2483
+
2484
+ 00:34:47.400 --> 00:34:49.640
2485
+ project, you can be 5 days late for two
2486
+
2487
+ 00:34:49.640 --> 00:34:50.830
2488
+ projects, and so on.
2489
+
2490
+ 00:34:51.810 --> 00:34:54.860
2491
+ And then there's a 5 point penalty per
2492
+
2493
+ 00:34:54.860 --> 00:34:55.900
2494
+ day after that.
2495
+
2496
+ 00:34:56.860 --> 00:34:58.510
2497
+ So for example, if you would have
2498
+
2499
+ 00:34:58.510 --> 00:35:02.422
2500
+ earned 110 points, but your five days
2501
+
2502
+ 00:35:02.422 --> 00:35:03.900
2503
+ late and you only had three late days
2504
+
2505
+ 00:35:03.900 --> 00:35:05.580
2506
+ left, then you would earn 100 points
2507
+
2508
+ 00:35:05.580 --> 00:35:06.090
2509
+ instead.
2510
+
2511
+ 00:35:07.170 --> 00:35:08.920
2512
+ And then the project.
2513
+
2514
+ 00:35:08.920 --> 00:35:11.945
2515
+ Every assignment, though, has to every
2516
+
2517
+ 00:35:11.945 --> 00:35:13.770
2518
+ homework has to be submitted within two
2519
+
2520
+ 00:35:13.770 --> 00:35:15.655
2521
+ weeks of the due date to receive any
2522
+
2523
+ 00:35:15.655 --> 00:35:15.930
2524
+ points.
2525
+
2526
+ 00:35:15.930 --> 00:35:19.114
2527
+ So you can't submit homework one like
2528
+
2529
+ 00:35:19.114 --> 00:35:20.786
2530
+ three weeks late and still get points
2531
+
2532
+ 00:35:20.786 --> 00:35:21.350
2533
+ for it.
2534
+
2535
+ 00:35:21.350 --> 00:35:23.201
2536
+ And part of the reason for that is that
2537
+
2538
+ 00:35:23.201 --> 00:35:25.769
2539
+ I don't want, I don't want any students
2540
+
2541
+ 00:35:25.769 --> 00:35:28.330
2542
+ to just get like chronically behind
2543
+
2544
+ 00:35:28.330 --> 00:35:29.583
2545
+ where you're submitting every
2546
+
2547
+ 00:35:29.583 --> 00:35:31.040
2548
+ assignment two or three weeks late,
2549
+
2550
+ 00:35:31.040 --> 00:35:32.050
2551
+ because I've seen that happen.
2552
+
2553
+ 00:35:32.050 --> 00:35:34.570
2554
+ And then I've had students build up
2555
+
2556
+ 00:35:34.570 --> 00:35:36.480
2557
+ like 60 late days by the end of this
2558
+
2559
+ 00:35:36.480 --> 00:35:37.280
2560
+ semester.
2561
+
2562
+ 00:35:37.930 --> 00:35:40.340
2563
+ So it's better just to move on if
2564
+
2565
+ 00:35:40.340 --> 00:35:41.330
2566
+ you're 2 weeks late.
2567
+
2568
+ 00:35:43.450 --> 00:35:44.460
2569
+
2570
+
2571
+ 00:35:46.510 --> 00:35:46.900
2572
+ Homework.
2573
+
2574
+ 00:35:48.990 --> 00:35:51.270
2575
+ The Final project I have not yet
2576
+
2577
+ 00:35:51.270 --> 00:35:53.690
2578
+ decided, but you definitely can't be
2579
+
2580
+ 00:35:53.690 --> 00:35:54.875
2581
+ the Final project.
2582
+
2583
+ 00:35:54.875 --> 00:35:57.890
2584
+ I might allow one or two late days, but
2585
+
2586
+ 00:35:57.890 --> 00:35:59.880
2587
+ probably not more than that because
2588
+
2589
+ 00:35:59.880 --> 00:36:02.340
2590
+ it's at the it's due on the last day of
2591
+
2592
+ 00:36:02.340 --> 00:36:04.030
2593
+ the semester, so there's not.
2594
+
2595
+ 00:36:04.030 --> 00:36:05.920
2596
+ So there needs to be time for grading
2597
+
2598
+ 00:36:05.920 --> 00:36:07.940
2599
+ and also time for you to do your exams.
2600
+
2601
+ 00:36:10.780 --> 00:36:13.350
2602
+ Alright, so Covid's Masks and sickness,
2603
+
2604
+ 00:36:13.350 --> 00:36:15.270
2605
+ so I do like it.
2606
+
2607
+ 00:36:15.270 --> 00:36:16.880
2608
+ If you come to I think it's a good idea
2609
+
2610
+ 00:36:16.880 --> 00:36:18.290
2611
+ to come to lecture whenever you can.
2612
+
2613
+ 00:36:18.290 --> 00:36:19.980
2614
+ I realize it's early in the morning for
2615
+
2616
+ 00:36:19.980 --> 00:36:22.770
2617
+ many people, but it's probably the best
2618
+
2619
+ 00:36:22.770 --> 00:36:26.865
2620
+ way to keep you on Schedule and even if
2621
+
2622
+ 00:36:26.865 --> 00:36:28.870
2623
+ you just come and.
2624
+
2625
+ 00:36:29.220 --> 00:36:32.179
2626
+ And our, I don't know, doing doing
2627
+
2628
+ 00:36:32.180 --> 00:36:32.881
2629
+ other work or whatever.
2630
+
2631
+ 00:36:32.881 --> 00:36:34.350
2632
+ At least you're at least you're like
2633
+
2634
+ 00:36:34.350 --> 00:36:35.750
2635
+ keeping tabs on what's going on in the
2636
+
2637
+ 00:36:35.750 --> 00:36:36.410
2638
+ course.
2639
+
2640
+ 00:36:36.970 --> 00:36:41.190
2641
+ But everything will be recorded, so
2642
+
2643
+ 00:36:41.190 --> 00:36:44.090
2644
+ it's you're also perfectly capable of
2645
+
2646
+ 00:36:44.090 --> 00:36:46.040
2647
+ catching up on anything that you missed
2648
+
2649
+ 00:36:46.040 --> 00:36:46.547
2650
+ online.
2651
+
2652
+ 00:36:46.547 --> 00:36:49.720
2653
+ So if you're well, do come to lectures
2654
+
2655
+ 00:36:49.720 --> 00:36:51.820
2656
+ and in office hours if you have
2657
+
2658
+ 00:36:51.820 --> 00:36:52.690
2659
+ something to discuss.
2660
+
2661
+ 00:36:53.610 --> 00:36:55.140
2662
+ Master optional.
2663
+
2664
+ 00:36:55.260 --> 00:36:58.440
2665
+ If you're sick, if you're sick, do you
2666
+
2667
+ 00:36:58.440 --> 00:37:00.120
2668
+ stay home because nobody else wants to
2669
+
2670
+ 00:37:00.120 --> 00:37:00.570
2671
+ get sick?
2672
+
2673
+ 00:37:01.560 --> 00:37:03.860
2674
+ You never need to show any proof of
2675
+
2676
+ 00:37:03.860 --> 00:37:06.910
2677
+ illness or you never need to ask me for
2678
+
2679
+ 00:37:06.910 --> 00:37:08.160
2680
+ permission to miss a class.
2681
+
2682
+ 00:37:08.160 --> 00:37:10.880
2683
+ You can just miss it and then you can
2684
+
2685
+ 00:37:10.880 --> 00:37:12.490
2686
+ watch the recording.
2687
+
2688
+ 00:37:12.490 --> 00:37:16.163
2689
+ So the lectures will be recorded so you
2690
+
2691
+ 00:37:16.163 --> 00:37:17.480
2692
+ can always catch up on them.
2693
+
2694
+ 00:37:17.480 --> 00:37:19.970
2695
+ As I was saying also the exams are
2696
+
2697
+ 00:37:19.970 --> 00:37:22.039
2698
+ planned to be on Prairie learn, so you
2699
+
2700
+ 00:37:22.040 --> 00:37:23.496
2701
+ would be able to take them at home.
2702
+
2703
+ 00:37:23.496 --> 00:37:24.874
2704
+ In fact, you won't be able to take them
2705
+
2706
+ 00:37:24.874 --> 00:37:27.080
2707
+ in the lecture, you will need to take
2708
+
2709
+ 00:37:27.080 --> 00:37:29.970
2710
+ them at home and the exams will be open
2711
+
2712
+ 00:37:29.970 --> 00:37:30.270
2713
+ book.
2714
+
2715
+ 00:37:32.600 --> 00:37:34.910
2716
+ Doesn't necessarily mean they're easy.
2717
+
2718
+ 00:37:36.460 --> 00:37:36.810
2719
+ Open.
2720
+
2721
+ 00:37:44.860 --> 00:37:46.420
2722
+ No, it's not.
2723
+
2724
+ 00:37:46.420 --> 00:37:48.810
2725
+ All right, so.
2726
+
2727
+ 00:37:48.880 --> 00:37:50.960
2728
+ For their homeworks, you will implement
2729
+
2730
+ 00:37:50.960 --> 00:37:52.870
2731
+ and apply machine learning methods in
2732
+
2733
+ 00:37:52.870 --> 00:37:53.810
2734
+ Jupiter notebooks.
2735
+
2736
+ 00:37:55.020 --> 00:37:55.895
2737
+ I'll show you.
2738
+
2739
+ 00:37:55.895 --> 00:37:57.640
2740
+ I'll go through the homework on the in
2741
+
2742
+ 00:37:57.640 --> 00:38:00.450
2743
+ the next lecture on Thursday, but
2744
+
2745
+ 00:38:00.450 --> 00:38:02.510
2746
+ you'll see that the basic structure is
2747
+
2748
+ 00:38:02.510 --> 00:38:04.630
2749
+ that you have a.
2750
+
2751
+ 00:38:05.260 --> 00:38:07.165
2752
+ You have like a main assignment page
2753
+
2754
+ 00:38:07.165 --> 00:38:10.220
2755
+ and then there's starter code which is
2756
+
2757
+ 00:38:10.220 --> 00:38:12.390
2758
+ pretty minimal but just enough to give
2759
+
2760
+ 00:38:12.390 --> 00:38:13.840
2761
+ some structure and to load the data
2762
+
2763
+ 00:38:13.840 --> 00:38:14.460
2764
+ that you need.
2765
+
2766
+ 00:38:15.220 --> 00:38:18.130
2767
+ And then there's a.
2768
+
2769
+ 00:38:18.760 --> 00:38:20.775
2770
+ So they're like places where you would
2771
+
2772
+ 00:38:20.775 --> 00:38:22.280
2773
+ where you would write the code for each
2774
+
2775
+ 00:38:22.280 --> 00:38:24.980
2776
+ section, and then there there's a
2777
+
2778
+ 00:38:24.980 --> 00:38:27.970
2779
+ report template which is template for
2780
+
2781
+ 00:38:27.970 --> 00:38:30.340
2782
+ reporting the results of your
2783
+
2784
+ 00:38:30.340 --> 00:38:31.040
2785
+ Algorithms.
2786
+
2787
+ 00:38:32.160 --> 00:38:34.300
2788
+ And then there's also like a tips page
2789
+
2790
+ 00:38:34.300 --> 00:38:36.660
2791
+ which has common code and some other
2792
+
2793
+ 00:38:36.660 --> 00:38:38.110
2794
+ suggestions about the assignment.
2795
+
2796
+ 00:38:41.270 --> 00:38:43.490
2797
+ Alright, so this is the main Learning
2798
+
2799
+ 00:38:43.490 --> 00:38:45.490
2800
+ resource, the website, I mean this is
2801
+
2802
+ 00:38:45.490 --> 00:38:47.170
2803
+ kind of the central repository of
2804
+
2805
+ 00:38:47.170 --> 00:38:48.230
2806
+ everything that you need.
2807
+
2808
+ 00:38:51.110 --> 00:38:53.190
2809
+ So that showed up there.
2810
+
2811
+ 00:38:53.190 --> 00:38:57.530
2812
+ OK, so you can find it from my home
2813
+
2814
+ 00:38:57.530 --> 00:38:58.030
2815
+ page.
2816
+
2817
+ 00:38:58.030 --> 00:38:59.427
2818
+ And also it's the.
2819
+
2820
+ 00:38:59.427 --> 00:39:01.405
2821
+ It's just like the standard location.
2822
+
2823
+ 00:39:01.405 --> 00:39:04.076
2824
+ So if you do like
2825
+
2826
+ 00:39:04.076 --> 00:39:06.909
2827
+ courses.endure.illinois.edu/CS 441.
2828
+
2829
+ 00:39:07.630 --> 00:39:09.320
2830
+ I think even that will go.
2831
+
2832
+ 00:39:09.320 --> 00:39:11.420
2833
+ Yeah, even that will just go to spring
2834
+
2835
+ 00:39:11.420 --> 00:39:12.090
2836
+ 2023.
2837
+
2838
+ 00:39:12.880 --> 00:39:15.895
2839
+ So you can see there's a Syllabus.
2840
+
2841
+ 00:39:15.895 --> 00:39:18.230
2842
+ There's obviously no lecture recordings
2843
+
2844
+ 00:39:18.230 --> 00:39:20.740
2845
+ yet, but once they should show up at
2846
+
2847
+ 00:39:20.740 --> 00:39:23.320
2848
+ this location on media space.
2849
+
2850
+ 00:39:24.770 --> 00:39:26.700
2851
+ There's the CampusWire.
2852
+
2853
+ 00:39:26.700 --> 00:39:28.160
2854
+ You can sign up with this code.
2855
+
2856
+ 00:39:28.160 --> 00:39:30.640
2857
+ I'll probably enroll everybody at some
2858
+
2859
+ 00:39:30.640 --> 00:39:32.575
2860
+ point this week that's registered for
2861
+
2862
+ 00:39:32.575 --> 00:39:33.190
2863
+ the class.
2864
+
2865
+ 00:39:34.350 --> 00:39:38.615
2866
+ The submission for Canvas, the
2867
+
2868
+ 00:39:38.615 --> 00:39:39.250
2869
+ Textbook.
2870
+
2871
+ 00:39:39.250 --> 00:39:42.400
2872
+ So I don't really teach out of a
2873
+
2874
+ 00:39:42.400 --> 00:39:46.160
2875
+ textbook, but this book is quite good.
2876
+
2877
+ 00:39:46.160 --> 00:39:47.620
2878
+ Applied machine Learning by David
2879
+
2880
+ 00:39:47.620 --> 00:39:49.350
2881
+ Forsyth, the professor here.
2882
+
2883
+ 00:39:49.350 --> 00:39:50.916
2884
+ He actually created the first version
2885
+
2886
+ 00:39:50.916 --> 00:39:54.685
2887
+ of this course and I do look at the
2888
+
2889
+ 00:39:54.685 --> 00:39:55.960
2890
+ Textbook to make sure I'm not
2891
+
2892
+ 00:39:55.960 --> 00:39:57.460
2893
+ forgetting anything really important.
2894
+
2895
+ 00:39:58.420 --> 00:40:01.650
2896
+ And it's useful to have like another
2897
+
2898
+ 00:40:01.650 --> 00:40:03.340
2899
+ perspective on some of the topics I'm
2900
+
2901
+ 00:40:03.340 --> 00:40:06.860
2902
+ teaching, or to have like more like
2903
+
2904
+ 00:40:06.860 --> 00:40:07.850
2905
+ background reading.
2906
+
2907
+ 00:40:08.930 --> 00:40:12.560
2908
+ So I'll this is the schedule of topics,
2909
+
2910
+ 00:40:12.560 --> 00:40:15.225
2911
+ so you can see that right now we're in
2912
+
2913
+ 00:40:15.225 --> 00:40:16.300
2914
+ the introduction.
2915
+
2916
+ 00:40:16.300 --> 00:40:18.750
2917
+ I've got the PowerPoint slides and the
2918
+
2919
+ 00:40:18.750 --> 00:40:19.375
2920
+ PDF here.
2921
+
2922
+ 00:40:19.375 --> 00:40:22.490
2923
+ I generally try to put them online
2924
+
2925
+ 00:40:22.490 --> 00:40:24.750
2926
+ before I teach, but I can't guarantee
2927
+
2928
+ 00:40:24.750 --> 00:40:25.250
2929
+ it.
2930
+
2931
+ 00:40:27.280 --> 00:40:29.940
2932
+ There is a I've got a bunch of
2933
+
2934
+ 00:40:29.940 --> 00:40:31.420
2935
+ tutorials here.
2936
+
2937
+ 00:40:31.420 --> 00:40:33.310
2938
+ These were originally created for the
2939
+
2940
+ 00:40:33.310 --> 00:40:35.050
2941
+ computational photography course, but
2942
+
2943
+ 00:40:35.050 --> 00:40:37.680
2944
+ they're pretty general, so one is on
2945
+
2946
+ 00:40:37.680 --> 00:40:40.630
2947
+ using Jupiter notebooks, one is on
2948
+
2949
+ 00:40:40.630 --> 00:40:42.590
2950
+ Numpy, and another is on linear
2951
+
2952
+ 00:40:42.590 --> 00:40:46.150
2953
+ algebra, and they're really good, so I
2954
+
2955
+ 00:40:46.150 --> 00:40:48.420
2956
+ would recommend watching those they
2957
+
2958
+ 00:40:48.420 --> 00:40:48.710
2959
+ have.
2960
+
2961
+ 00:40:48.710 --> 00:40:51.860
2962
+ It's a video, but some of them, like
2963
+
2964
+ 00:40:51.860 --> 00:40:53.770
2965
+ this one for Jupiter notebook, also
2966
+
2967
+ 00:40:53.770 --> 00:40:55.780
2968
+ comes with a notebook and they'll like
2969
+
2970
+ 00:40:55.780 --> 00:40:57.380
2971
+ pause and give you time to try things
2972
+
2973
+ 00:40:57.380 --> 00:40:58.050
2974
+ out yourself.
2975
+
2976
+ 00:40:59.110 --> 00:41:00.540
2977
+ So those are worth reviewing,
2978
+
2979
+ 00:41:00.540 --> 00:41:02.570
2980
+ especially if you're not familiar with
2981
+
2982
+ 00:41:02.570 --> 00:41:04.230
2983
+ Jupiter notebooks, if you're not
2984
+
2985
+ 00:41:04.230 --> 00:41:08.290
2986
+ familiar with Python, or if you've feel
2987
+
2988
+ 00:41:08.290 --> 00:41:09.620
2989
+ like you need a refresher on linear
2990
+
2991
+ 00:41:09.620 --> 00:41:10.130
2992
+ algebra.
2993
+
2994
+ 00:41:11.900 --> 00:41:13.900
2995
+ The Assignments are linked to here, so
2996
+
2997
+ 00:41:13.900 --> 00:41:15.960
2998
+ if you click on that you can it should
2999
+
3000
+ 00:41:15.960 --> 00:41:17.170
3001
+ take you to homework one.
3002
+
3003
+ 00:41:17.890 --> 00:41:21.480
3004
+ And you can check that out if you're
3005
+
3006
+ 00:41:21.480 --> 00:41:22.770
3007
+ interested.
3008
+
3009
+ 00:41:22.770 --> 00:41:24.770
3010
+ Like I said, I'm going to review that
3011
+
3012
+ 00:41:24.770 --> 00:41:26.082
3013
+ on Thursday.
3014
+
3015
+ 00:41:26.082 --> 00:41:30.445
3016
+ And that first homework mainly is the
3017
+
3018
+ 00:41:30.445 --> 00:41:32.450
3019
+ that covers the topics that are taught
3020
+
3021
+ 00:41:32.450 --> 00:41:33.650
3022
+ in the first three lectures.
3023
+
3024
+ 00:41:33.650 --> 00:41:34.830
3025
+ KNN maybes.
3026
+
3027
+ 00:41:35.460 --> 00:41:37.860
3028
+ And Linear Logistic Regression.
3029
+
3030
+ 00:41:39.990 --> 00:41:41.250
3031
+ So.
3032
+
3033
+ 00:41:42.260 --> 00:41:42.630
3034
+ Yes.
3035
+
3036
+ 00:41:42.630 --> 00:41:43.770
3037
+ So the other thing.
3038
+
3039
+ 00:41:45.510 --> 00:41:46.440
3040
+ Show you the Syllabus.
3041
+
3042
+ 00:41:46.440 --> 00:41:47.520
3043
+ So here's the Syllabus.
3044
+
3045
+ 00:41:48.430 --> 00:41:49.755
3046
+ Do you read it on your own?
3047
+
3048
+ 00:41:49.755 --> 00:41:52.660
3049
+ You can see more or less the same
3050
+
3051
+ 00:41:52.660 --> 00:41:55.240
3052
+ information that I just gave you, but
3053
+
3054
+ 00:41:55.240 --> 00:41:57.460
3055
+ it has the, I guess a little bit more
3056
+
3057
+ 00:41:57.460 --> 00:41:58.322
3058
+ detail on some things.
3059
+
3060
+ 00:41:58.322 --> 00:41:59.950
3061
+ It has the greeting equations.
3062
+
3063
+ 00:42:00.880 --> 00:42:03.680
3064
+ So the also has grade Thresholds, so.
3065
+
3066
+ 00:42:04.570 --> 00:42:06.001
3067
+ If you reach these graded Thresholds
3068
+
3069
+ 00:42:06.001 --> 00:42:07.507
3070
+ then you're guaranteed that grade.
3071
+
3072
+ 00:42:07.507 --> 00:42:10.219
3073
+ So if you get like a 97 you will
3074
+
3075
+ 00:42:10.220 --> 00:42:11.480
3076
+ definitely get an A plus.
3077
+
3078
+ 00:42:11.480 --> 00:42:13.940
3079
+ I generally will look at the grade
3080
+
3081
+ 00:42:13.940 --> 00:42:16.560
3082
+ distribution at the end of the semester
3083
+
3084
+ 00:42:16.560 --> 00:42:19.790
3085
+ and if it seems warranted then I might
3086
+
3087
+ 00:42:19.790 --> 00:42:21.710
3088
+ lower the Thresholds but I won't raise
3089
+
3090
+ 00:42:21.710 --> 00:42:22.000
3091
+ them.
3092
+
3093
+ 00:42:22.000 --> 00:42:24.147
3094
+ So if you get like a 90, you're
3095
+
3096
+ 00:42:24.147 --> 00:42:24.953
3097
+ guaranteed a minus.
3098
+
3099
+ 00:42:24.953 --> 00:42:27.209
3100
+ It might be that if you get an 89 you
3101
+
3102
+ 00:42:27.210 --> 00:42:30.444
3103
+ could also get an A minus, but I it
3104
+
3105
+ 00:42:30.444 --> 00:42:31.205
3106
+ depends on.
3107
+
3108
+ 00:42:31.205 --> 00:42:34.110
3109
+ It depends on like I have to see.
3110
+
3111
+ 00:42:34.160 --> 00:42:36.130
3112
+ Have to at the end of the semester so I
3113
+
3114
+ 00:42:36.130 --> 00:42:38.080
3115
+ don't curve any individual Assignments
3116
+
3117
+ 00:42:38.080 --> 00:42:38.500
3118
+ I do.
3119
+
3120
+ 00:42:38.500 --> 00:42:39.820
3121
+ I can curve.
3122
+
3123
+ 00:42:40.770 --> 00:42:43.670
3124
+ The final grades, but usually it will
3125
+
3126
+ 00:42:43.670 --> 00:42:45.480
3127
+ not change very much.
3128
+
3129
+ 00:42:45.480 --> 00:42:47.780
3130
+ So just think of these as your targets.
3131
+
3132
+ 00:42:49.530 --> 00:42:51.320
3133
+ Late policy explained.
3134
+
3135
+ 00:42:52.810 --> 00:42:55.650
3136
+ I think all of this, yeah.
3137
+
3138
+ 00:42:55.650 --> 00:42:59.340
3139
+ So I guess it's worth noting that this
3140
+
3141
+ 00:42:59.340 --> 00:43:01.420
3142
+ is the first time I've taught this
3143
+
3144
+ 00:43:01.420 --> 00:43:04.740
3145
+ course, and so I.
3146
+
3147
+ 00:43:05.490 --> 00:43:08.660
3148
+ I'm I wanna keep some flexibility
3149
+
3150
+ 00:43:08.660 --> 00:43:09.255
3151
+ options open.
3152
+
3153
+ 00:43:09.255 --> 00:43:11.060
3154
+ I'm going to be soliciting feedback at
3155
+
3156
+ 00:43:11.060 --> 00:43:11.850
3157
+ various points.
3158
+
3159
+ 00:43:12.670 --> 00:43:16.480
3160
+ I may course correct, so I'll do so in
3161
+
3162
+ 00:43:16.480 --> 00:43:18.530
3163
+ a way that's not disruptive and give
3164
+
3165
+ 00:43:18.530 --> 00:43:20.320
3166
+ you as much notice as possible.
3167
+
3168
+ 00:43:20.320 --> 00:43:22.050
3169
+ But I don't want to set.
3170
+
3171
+ 00:43:22.050 --> 00:43:23.410
3172
+ I don't want to.
3173
+
3174
+ 00:43:24.670 --> 00:43:26.680
3175
+ To be completely inflexible just so
3176
+
3177
+ 00:43:26.680 --> 00:43:28.820
3178
+ that I can make improvements that may
3179
+
3180
+ 00:43:28.820 --> 00:43:30.110
3181
+ benefit your experience.
3182
+
3183
+ 00:43:32.170 --> 00:43:35.470
3184
+ Alright, so yeah, that's enough of
3185
+
3186
+ 00:43:35.470 --> 00:43:36.182
3187
+ reviewing this.
3188
+
3189
+ 00:43:36.182 --> 00:43:38.120
3190
+ So you should read all of this stuff,
3191
+
3192
+ 00:43:38.120 --> 00:43:40.620
3193
+ but I don't need to do it right now.
3194
+
3195
+ 00:43:42.800 --> 00:43:44.400
3196
+ Alright, I think I talked about that
3197
+
3198
+ 00:43:44.400 --> 00:43:45.310
3199
+ stuff.
3200
+
3201
+ 00:43:45.310 --> 00:43:50.900
3202
+ OK, so the office hours, I do have them
3203
+
3204
+ 00:43:50.900 --> 00:43:53.270
3205
+ ready, but I'll create a post on
3206
+
3207
+ 00:43:53.270 --> 00:43:55.120
3208
+ CampusWire that will tell you what the
3209
+
3210
+ 00:43:55.120 --> 00:43:56.580
3211
+ office hours and pin them.
3212
+
3213
+ 00:43:56.580 --> 00:43:58.340
3214
+ The office hours will start next week.
3215
+
3216
+ 00:43:59.390 --> 00:44:01.380
3217
+ And then, like I said, the reading.
3218
+
3219
+ 00:44:01.380 --> 00:44:03.010
3220
+ The main Readings are the applied
3221
+
3222
+ 00:44:03.010 --> 00:44:05.260
3223
+ machine learning book by David Forsyth.
3224
+
3225
+ 00:44:09.660 --> 00:44:12.370
3226
+ Alright, so Academic Integrity.
3227
+
3228
+ 00:44:12.470 --> 00:44:13.190
3229
+
3230
+
3231
+ 00:44:14.040 --> 00:44:15.860
3232
+ It's OK to discuss your homework with
3233
+
3234
+ 00:44:15.860 --> 00:44:16.750
3235
+ classmates.
3236
+
3237
+ 00:44:16.750 --> 00:44:20.780
3238
+ And actually, if you and somebody else
3239
+
3240
+ 00:44:20.780 --> 00:44:23.100
3241
+ finish your homework, I would actually
3242
+
3243
+ 00:44:23.100 --> 00:44:26.585
3244
+ encourage you to like to like, discuss
3245
+
3246
+ 00:44:26.585 --> 00:44:28.304
3247
+ like what were your results?
3248
+
3249
+ 00:44:28.304 --> 00:44:30.520
3250
+ And if your results don't match, then
3251
+
3252
+ 00:44:30.520 --> 00:44:31.880
3253
+ find out what they did.
3254
+
3255
+ 00:44:31.880 --> 00:44:34.570
3256
+ I'm OK with that, but if you haven't
3257
+
3258
+ 00:44:34.570 --> 00:44:36.330
3259
+ done it yet, don't like look at their
3260
+
3261
+ 00:44:36.330 --> 00:44:39.130
3262
+ code so that you can do the exact same
3263
+
3264
+ 00:44:39.130 --> 00:44:40.316
3265
+ thing that they did.
3266
+
3267
+ 00:44:40.316 --> 00:44:41.990
3268
+ So I think, like, there's actually a
3269
+
3270
+ 00:44:41.990 --> 00:44:43.850
3271
+ lot of educational value in learning
3272
+
3273
+ 00:44:43.850 --> 00:44:45.180
3274
+ from other students and.
3275
+
3276
+ 00:44:45.890 --> 00:44:47.392
3277
+ Sometimes there's going to be more than
3278
+
3279
+ 00:44:47.392 --> 00:44:49.470
3280
+ one way to implement something, and one
3281
+
3282
+ 00:44:49.470 --> 00:44:50.840
3283
+ way will lead to better results than
3284
+
3285
+ 00:44:50.840 --> 00:44:51.300
3286
+ another.
3287
+
3288
+ 00:44:51.950 --> 00:44:55.280
3289
+ And it's worth finding out what that
3290
+
3291
+ 00:44:55.280 --> 00:44:56.537
3292
+ better way is.
3293
+
3294
+ 00:44:56.537 --> 00:44:59.980
3295
+ But you should mainly do things
3296
+
3297
+ 00:44:59.980 --> 00:45:04.070
3298
+ independently for the Assignments.
3299
+
3300
+ 00:45:04.070 --> 00:45:05.710
3301
+ Like I said, for the Final project you
3302
+
3303
+ 00:45:05.710 --> 00:45:07.300
3304
+ can work in groups, but for the regular
3305
+
3306
+ 00:45:07.300 --> 00:45:08.570
3307
+ Assignments it should be mostly
3308
+
3309
+ 00:45:08.570 --> 00:45:09.310
3310
+ independent work.
3311
+
3312
+ 00:45:09.950 --> 00:45:12.000
3313
+ And any consultation with others is
3314
+
3315
+ 00:45:12.000 --> 00:45:14.220
3316
+ should be aimed at further enhancing
3317
+
3318
+ 00:45:14.220 --> 00:45:16.700
3319
+ your what you learn rather than
3320
+
3321
+ 00:45:16.700 --> 00:45:17.800
3322
+ bypassing it.
3323
+
3324
+ 00:45:19.730 --> 00:45:21.322
3325
+ You can look at Stack Overflow.
3326
+
3327
+ 00:45:21.322 --> 00:45:23.036
3328
+ You can get ideas from online.
3329
+
3330
+ 00:45:23.036 --> 00:45:25.700
3331
+ At the end of the template there's a
3332
+
3333
+ 00:45:25.700 --> 00:45:27.940
3334
+ section where you say what other
3335
+
3336
+ 00:45:27.940 --> 00:45:29.340
3337
+ sources you have.
3338
+
3339
+ 00:45:29.340 --> 00:45:32.031
3340
+ And so if you talk to anybody about the
3341
+
3342
+ 00:45:32.031 --> 00:45:33.234
3343
+ course, I mean not the course.
3344
+
3345
+ 00:45:33.234 --> 00:45:34.439
3346
+ If you talk to anybody about the
3347
+
3348
+ 00:45:34.440 --> 00:45:36.132
3349
+ assignment or looked at Stack Overflow
3350
+
3351
+ 00:45:36.132 --> 00:45:38.410
3352
+ or whatever, you should just list those
3353
+
3354
+ 00:45:38.410 --> 00:45:38.960
3355
+ things there.
3356
+
3357
+ 00:45:38.960 --> 00:45:39.280
3358
+ So.
3359
+
3360
+ 00:45:40.070 --> 00:45:42.600
3361
+ As you're doing the assignment if you.
3362
+
3363
+ 00:45:42.810 --> 00:45:44.140
3364
+ If you have like.
3365
+
3366
+ 00:45:45.780 --> 00:45:48.190
3367
+ If you end up using other resources,
3368
+
3369
+ 00:45:48.190 --> 00:45:49.840
3370
+ just write them down South that you
3371
+
3372
+ 00:45:49.840 --> 00:45:50.870
3373
+ remember to put them there.
3374
+
3375
+ 00:45:52.500 --> 00:45:55.470
3376
+ So you shouldn't like copy any code
3377
+
3378
+ 00:45:55.470 --> 00:45:56.570
3379
+ from anybody.
3380
+
3381
+ 00:45:56.570 --> 00:45:58.530
3382
+ So you should never be like claiming
3383
+
3384
+ 00:45:58.530 --> 00:46:00.250
3385
+ credit for something that somebody else
3386
+
3387
+ 00:46:00.250 --> 00:46:01.510
3388
+ wrote, even if you typed it out
3389
+
3390
+ 00:46:01.510 --> 00:46:02.880
3391
+ yourself.
3392
+
3393
+ 00:46:02.880 --> 00:46:05.050
3394
+ And you should not use external
3395
+
3396
+ 00:46:05.050 --> 00:46:06.620
3397
+ resources but without acknowledging
3398
+
3399
+ 00:46:06.620 --> 00:46:06.840
3400
+ them.
3401
+
3402
+ 00:46:07.640 --> 00:46:08.950
3403
+ So if you're not sure if something's
3404
+
3405
+ 00:46:08.950 --> 00:46:11.900
3406
+ OK, you can just ask and as long as you
3407
+
3408
+ 00:46:11.900 --> 00:46:14.279
3409
+ acknowledge all your sources then you
3410
+
3411
+ 00:46:14.279 --> 00:46:16.230
3412
+ can then you're safe.
3413
+
3414
+ 00:46:16.230 --> 00:46:16.830
3415
+ So.
3416
+
3417
+ 00:46:17.610 --> 00:46:19.865
3418
+ Even if even if what you did is
3419
+
3420
+ 00:46:19.865 --> 00:46:22.685
3421
+ something that I would consider to be
3422
+
3423
+ 00:46:22.685 --> 00:46:25.524
3424
+ like 2 too much copying or something,
3425
+
3426
+ 00:46:25.524 --> 00:46:27.290
3427
+ if you just if you disclose it and
3428
+
3429
+ 00:46:27.290 --> 00:46:28.410
3430
+ you're acknowledgements, it's not
3431
+
3432
+ 00:46:28.410 --> 00:46:32.031
3433
+ cheating, it's just not would not be
3434
+
3435
+ 00:46:32.031 --> 00:46:33.840
3436
+ like doing enough I guess.
3437
+
3438
+ 00:46:36.840 --> 00:46:38.110
3439
+ So.
3440
+
3441
+ 00:46:39.080 --> 00:46:39.840
3442
+ Alright.
3443
+
3444
+ 00:46:39.840 --> 00:46:42.800
3445
+ So yeah, in terms of Prerequisites, so
3446
+
3447
+ 00:46:42.800 --> 00:46:44.380
3448
+ I'm going to assume that you've had
3449
+
3450
+ 00:46:44.380 --> 00:46:46.140
3451
+ some kind of probability and statistics
3452
+
3453
+ 00:46:46.140 --> 00:46:46.990
3454
+ before.
3455
+
3456
+ 00:46:46.990 --> 00:46:49.240
3457
+ I will review it a little bit when I
3458
+
3459
+ 00:46:49.240 --> 00:46:52.790
3460
+ talk about Naive Bayes, but it's really
3461
+
3462
+ 00:46:52.790 --> 00:46:56.060
3463
+ like a 20 or 30 minute review to what
3464
+
3465
+ 00:46:56.060 --> 00:46:57.990
3466
+ would normally be a course.
3467
+
3468
+ 00:46:57.990 --> 00:47:01.019
3469
+ So it's not meant to replace replace
3470
+
3471
+ 00:47:01.020 --> 00:47:03.130
3472
+ having taken probably probability or
3473
+
3474
+ 00:47:03.130 --> 00:47:04.905
3475
+ statistics and that's not supposed to
3476
+
3477
+ 00:47:04.905 --> 00:47:07.890
3478
+ be stages, it's stats I think.
3479
+
3480
+ 00:47:08.460 --> 00:47:11.610
3481
+ And the linear algebra.
3482
+
3483
+ 00:47:11.610 --> 00:47:13.410
3484
+ So again, I'm going to assume that you
3485
+
3486
+ 00:47:13.410 --> 00:47:15.830
3487
+ things like matrix multiplication, what
3488
+
3489
+ 00:47:15.830 --> 00:47:18.070
3490
+ inverses, stuff like that.
3491
+
3492
+ 00:47:19.280 --> 00:47:23.030
3493
+ If not, if not, then you should
3494
+
3495
+ 00:47:23.030 --> 00:47:26.010
3496
+ probably learn it first, but you can
3497
+
3498
+ 00:47:26.010 --> 00:47:27.590
3499
+ review it using the tutorial.
3500
+
3501
+ 00:47:28.450 --> 00:47:30.930
3502
+ And then also assume some knowledge of
3503
+
3504
+ 00:47:30.930 --> 00:47:33.230
3505
+ calculus, like if I take a derivative,
3506
+
3507
+ 00:47:33.230 --> 00:47:33.620
3508
+ what is?
3509
+
3510
+ 00:47:33.620 --> 00:47:35.280
3511
+ And you kind of know you how to take
3512
+
3513
+ 00:47:35.280 --> 00:47:36.480
3514
+ derivatives yourself.
3515
+
3516
+ 00:47:37.530 --> 00:47:40.490
3517
+ And if you have experience with Python,
3518
+
3519
+ 00:47:40.490 --> 00:47:41.990
3520
+ that will help, because we'll be doing
3521
+
3522
+ 00:47:41.990 --> 00:47:44.440
3523
+ the assignments in Python, but it's not
3524
+
3525
+ 00:47:44.440 --> 00:47:46.920
3526
+ necessary, I mean I myself.
3527
+
3528
+ 00:47:48.780 --> 00:47:49.820
3529
+ Don't.
3530
+
3531
+ 00:47:49.820 --> 00:47:52.490
3532
+ I've used Python a bit.
3533
+
3534
+ 00:47:52.560 --> 00:47:55.750
3535
+ And I find it pretty easy to do the
3536
+
3537
+ 00:47:55.750 --> 00:47:58.590
3538
+ coding, but I'm not like an expert in
3539
+
3540
+ 00:47:58.590 --> 00:48:00.404
3541
+ Python by any means, so it's not.
3542
+
3543
+ 00:48:00.404 --> 00:48:02.016
3544
+ It's not extremely challenging coding
3545
+
3546
+ 00:48:02.016 --> 00:48:03.257
3547
+ in my opinion.
3548
+
3549
+ 00:48:03.257 --> 00:48:06.823
3550
+ So if you're generally capable of
3551
+
3552
+ 00:48:06.823 --> 00:48:08.170
3553
+ coding, then you're capable of picking
3554
+
3555
+ 00:48:08.170 --> 00:48:09.890
3556
+ up Python And doing it, but if you
3557
+
3558
+ 00:48:09.890 --> 00:48:10.810
3559
+ already know, it will help.
3560
+
3561
+ 00:48:12.290 --> 00:48:14.090
3562
+ And then like I said, if watch the
3563
+
3564
+ 00:48:14.090 --> 00:48:16.820
3565
+ tutorials if you would like to review
3566
+
3567
+ 00:48:16.820 --> 00:48:17.740
3568
+ those concepts.
3569
+
3570
+ 00:48:20.990 --> 00:48:24.290
3571
+ So I also want to just briefly mention
3572
+
3573
+ 00:48:24.290 --> 00:48:25.840
3574
+ how this course is different from some
3575
+
3576
+ 00:48:25.840 --> 00:48:27.280
3577
+ others, because that's a really common
3578
+
3579
+ 00:48:27.280 --> 00:48:28.465
3580
+ question.
3581
+
3582
+ 00:48:28.465 --> 00:48:32.320
3583
+ So one of the other main courses is
3584
+
3585
+ 00:48:32.320 --> 00:48:34.610
3586
+ 446, which is just called machine
3587
+
3588
+ 00:48:34.610 --> 00:48:35.120
3589
+ learning.
3590
+
3591
+ 00:48:36.080 --> 00:48:38.490
3592
+ So this course, when I say this course
3593
+
3594
+ 00:48:38.490 --> 00:48:40.180
3595
+ here, I mean this one that we're in
3596
+
3597
+ 00:48:40.180 --> 00:48:43.078
3598
+ right now, 441 this course provides a
3599
+
3600
+ 00:48:43.078 --> 00:48:45.160
3601
+ foundation for ML practice, while I
3602
+
3603
+ 00:48:45.160 --> 00:48:47.370
3604
+ would say 446 provides more of a
3605
+
3606
+ 00:48:47.370 --> 00:48:48.700
3607
+ foundation for machine learning
3608
+
3609
+ 00:48:48.700 --> 00:48:49.630
3610
+ research.
3611
+
3612
+ 00:48:49.630 --> 00:48:52.710
3613
+ And so compared to 446, we will have
3614
+
3615
+ 00:48:52.710 --> 00:48:55.660
3616
+ less theory, fewer derivations, less
3617
+
3618
+ 00:48:55.660 --> 00:48:59.440
3619
+ focus on optimization and more focus on
3620
+
3621
+ 00:48:59.440 --> 00:49:01.610
3622
+ how you represent things on data
3623
+
3624
+ 00:49:01.610 --> 00:49:03.300
3625
+ representations and developing
3626
+
3627
+ 00:49:03.300 --> 00:49:05.510
3628
+ applications and.
3629
+
3630
+ 00:49:05.570 --> 00:49:07.140
3631
+ Application examples.
3632
+
3633
+ 00:49:07.140 --> 00:49:08.620
3634
+ So it doesn't mean that we don't have
3635
+
3636
+ 00:49:08.620 --> 00:49:10.630
3637
+ any theory, but it's all everything in
3638
+
3639
+ 00:49:10.630 --> 00:49:12.790
3640
+ the course is geared towards making you
3641
+
3642
+ 00:49:12.790 --> 00:49:15.860
3643
+ a good machine learning practitioner
3644
+
3645
+ 00:49:15.860 --> 00:49:19.035
3646
+ rather than making you a good machine
3647
+
3648
+ 00:49:19.035 --> 00:49:20.340
3649
+ learning researcher.
3650
+
3651
+ 00:49:20.340 --> 00:49:22.260
3652
+ There are just different focuses.
3653
+
3654
+ 00:49:23.090 --> 00:49:24.580
3655
+ Also, if you look at the syllabus for
3656
+
3657
+ 00:49:24.580 --> 00:49:27.090
3658
+ 446, you'll see that there are some
3659
+
3660
+ 00:49:27.090 --> 00:49:28.770
3661
+ topics that are in common and then
3662
+
3663
+ 00:49:28.770 --> 00:49:30.540
3664
+ there's others that are different so.
3665
+
3666
+ 00:49:31.660 --> 00:49:33.740
3667
+ But you can just, like look at doing an
3668
+
3669
+ 00:49:33.740 --> 00:49:36.490
3670
+ AB comparison on the syllabi once if
3671
+
3672
+ 00:49:36.490 --> 00:49:38.390
3673
+ the 446 Syllabus is available.
3674
+
3675
+ 00:49:38.390 --> 00:49:40.700
3676
+ There's also an online version of this
3677
+
3678
+ 00:49:40.700 --> 00:49:42.640
3679
+ course that's currently taught by Marco
3680
+
3681
+ 00:49:42.640 --> 00:49:43.160
3682
+ Morales.
3683
+
3684
+ 00:49:45.060 --> 00:49:48.220
3685
+ This is a complete redesign for this
3686
+
3687
+ 00:49:48.220 --> 00:49:48.610
3688
+ semester.
3689
+
3690
+ 00:49:48.610 --> 00:49:52.500
3691
+ My version is a total redesign, so they
3692
+
3693
+ 00:49:52.500 --> 00:49:54.680
3694
+ cover similar topics but in different
3695
+
3696
+ 00:49:54.680 --> 00:49:55.170
3697
+ ways.
3698
+
3699
+ 00:49:57.340 --> 00:50:00.820
3700
+ Assignment wise, this course has fewer
3701
+
3702
+ 00:50:00.820 --> 00:50:02.700
3703
+ and larger homeworks that are a little
3704
+
3705
+ 00:50:02.700 --> 00:50:04.550
3706
+ bit more open-ended.
3707
+
3708
+ 00:50:06.450 --> 00:50:08.780
3709
+ And also as a Final project and exams,
3710
+
3711
+ 00:50:08.780 --> 00:50:10.820
3712
+ while the online version, at least the
3713
+
3714
+ 00:50:10.820 --> 00:50:13.780
3715
+ last one I saw has like 11 homeworks
3716
+
3717
+ 00:50:13.780 --> 00:50:16.210
3718
+ that are a little bit more scripted and
3719
+
3720
+ 00:50:16.210 --> 00:50:20.100
3721
+ it has quizzes and it's just a.
3722
+
3723
+ 00:50:21.320 --> 00:50:23.050
3724
+ Very different in the kind of
3725
+
3726
+ 00:50:23.050 --> 00:50:23.500
3727
+ structure.
3728
+
3729
+ 00:50:25.420 --> 00:50:26.090
3730
+ And.
3731
+
3732
+ 00:50:27.300 --> 00:50:29.720
3733
+ And I would say that compared to the
3734
+
3735
+ 00:50:29.720 --> 00:50:31.660
3736
+ online one, the one that I'm teaching
3737
+
3738
+ 00:50:31.660 --> 00:50:33.870
3739
+ now focuses more on a conceptual
3740
+
3741
+ 00:50:33.870 --> 00:50:37.300
3742
+ understanding of the techniques and on
3743
+
3744
+ 00:50:37.300 --> 00:50:38.850
3745
+ how machine learning is used today.
3746
+
3747
+ 00:50:40.730 --> 00:50:43.440
3748
+ And then one more example is there's a
3749
+
3750
+ 00:50:43.440 --> 00:50:45.005
3751
+ deep Learning course for computer
3752
+
3753
+ 00:50:45.005 --> 00:50:45.710
3754
+ vision.
3755
+
3756
+ 00:50:45.710 --> 00:50:48.566
3757
+ So that's focused on, as it says, deep
3758
+
3759
+ 00:50:48.566 --> 00:50:50.400
3760
+ learning and computer vision, where
3761
+
3762
+ 00:50:50.400 --> 00:50:52.916
3763
+ this course is not only focused on deep
3764
+
3765
+ 00:50:52.916 --> 00:50:54.850
3766
+ learning, I teach a variety of
3767
+
3768
+ 00:50:54.850 --> 00:50:56.955
3769
+ different techniques, including deep
3770
+
3771
+ 00:50:56.955 --> 00:50:58.090
3772
+ learning to some extent.
3773
+
3774
+ 00:50:58.770 --> 00:51:01.850
3775
+ And also talk about a broader variety
3776
+
3777
+ 00:51:01.850 --> 00:51:02.890
3778
+ of Application Domains.
3779
+
3780
+ 00:51:04.570 --> 00:51:06.482
3781
+ So you may be wondering whether you
3782
+
3783
+ 00:51:06.482 --> 00:51:07.310
3784
+ take the course.
3785
+
3786
+ 00:51:07.310 --> 00:51:09.029
3787
+ I would say you should take the course
3788
+
3789
+ 00:51:09.030 --> 00:51:10.150
3790
+ if you want to learn how to apply
3791
+
3792
+ 00:51:10.150 --> 00:51:10.760
3793
+ machine learning.
3794
+
3795
+ 00:51:11.810 --> 00:51:13.626
3796
+ If you like coding based homeworks and
3797
+
3798
+ 00:51:13.626 --> 00:51:15.280
3799
+ you're also OK with math, then you
3800
+
3801
+ 00:51:15.280 --> 00:51:16.863
3802
+ might want to take the course.
3803
+
3804
+ 00:51:16.863 --> 00:51:20.210
3805
+ Also, you need to be willing to spend a
3806
+
3807
+ 00:51:20.210 --> 00:51:21.930
3808
+ significant amount of time, so I would
3809
+
3810
+ 00:51:21.930 --> 00:51:23.700
3811
+ say it would probably take 10 to 12
3812
+
3813
+ 00:51:23.700 --> 00:51:24.760
3814
+ hours per week.
3815
+
3816
+ 00:51:24.760 --> 00:51:28.103
3817
+ If you don't know if you're catching up
3818
+
3819
+ 00:51:28.103 --> 00:51:29.610
3820
+ on a lot of things then it could take
3821
+
3822
+ 00:51:29.610 --> 00:51:31.806
3823
+ even longer or depending on how fast of
3824
+
3825
+ 00:51:31.806 --> 00:51:33.290
3826
+ a programmer you are and things like
3827
+
3828
+ 00:51:33.290 --> 00:51:33.840
3829
+ that.
3830
+
3831
+ 00:51:33.840 --> 00:51:37.470
3832
+ But this is my estimate, so it is a
3833
+
3834
+ 00:51:37.470 --> 00:51:38.450
3835
+ it's not a light course.
3836
+
3837
+ 00:51:40.230 --> 00:51:42.640
3838
+ Don't take the course if so.
3839
+
3840
+ 00:51:42.640 --> 00:51:43.790
3841
+ If you want a more theoretical
3842
+
3843
+ 00:51:43.790 --> 00:51:46.070
3844
+ background, 446 would be more
3845
+
3846
+ 00:51:46.070 --> 00:51:47.680
3847
+ specifically for you.
3848
+
3849
+ 00:51:47.680 --> 00:51:49.590
3850
+ I think there could be value in taking
3851
+
3852
+ 00:51:49.590 --> 00:51:50.300
3853
+ both of them.
3854
+
3855
+ 00:51:50.300 --> 00:51:52.630
3856
+ When was I took like a big variety of
3857
+
3858
+ 00:51:52.630 --> 00:51:55.000
3859
+ machine learning courses and I never
3860
+
3861
+ 00:51:55.000 --> 00:51:56.400
3862
+ really minded if they had overlap
3863
+
3864
+ 00:51:56.400 --> 00:51:57.770
3865
+ because I always found it useful to
3866
+
3867
+ 00:51:57.770 --> 00:51:59.650
3868
+ revisit a topic and to get different
3869
+
3870
+ 00:51:59.650 --> 00:52:00.990
3871
+ perspectives on it.
3872
+
3873
+ 00:52:00.990 --> 00:52:03.460
3874
+ But 446 is more theory oriented.
3875
+
3876
+ 00:52:04.610 --> 00:52:06.770
3877
+ And if you want to focus on a single
3878
+
3879
+ 00:52:06.770 --> 00:52:08.460
3880
+ Application Domain, again there's more
3881
+
3882
+ 00:52:08.460 --> 00:52:10.680
3883
+ focused courses for that, like a Vision
3884
+
3885
+ 00:52:10.680 --> 00:52:13.020
3886
+ or NLP or special topics course.
3887
+
3888
+ 00:52:13.680 --> 00:52:15.650
3889
+ And if you want an easy A, this is also
3890
+
3891
+ 00:52:15.650 --> 00:52:18.160
3892
+ probably not the right course.
3893
+
3894
+ 00:52:18.160 --> 00:52:20.022
3895
+ It is possible for everybody to get in
3896
+
3897
+ 00:52:20.022 --> 00:52:21.790
3898
+ a if you were to.
3899
+
3900
+ 00:52:22.180 --> 00:52:23.970
3901
+ To do well on the Assignments and
3902
+
3903
+ 00:52:23.970 --> 00:52:26.750
3904
+ exams, but it's definitely not going to
3905
+
3906
+ 00:52:26.750 --> 00:52:27.210
3907
+ be easy.
3908
+
3909
+ 00:52:29.920 --> 00:52:32.090
3910
+ And as I mentioned earlier, your
3911
+
3912
+ 00:52:32.090 --> 00:52:33.240
3913
+ feedback is welcome.
3914
+
3915
+ 00:52:33.240 --> 00:52:35.715
3916
+ So I will occasionally solicit feedback
3917
+
3918
+ 00:52:35.715 --> 00:52:37.590
3919
+ and if you respond, that would help me.
3920
+
3921
+ 00:52:38.580 --> 00:52:40.005
3922
+ You can always talk to me after class
3923
+
3924
+ 00:52:40.005 --> 00:52:41.860
3925
+ or send me a message on CampusWire or
3926
+
3927
+ 00:52:41.860 --> 00:52:43.530
3928
+ come to my office hours.
3929
+
3930
+ 00:52:43.630 --> 00:52:47.530
3931
+ And my philosophy in teaching is to be
3932
+
3933
+ 00:52:47.530 --> 00:52:48.715
3934
+ a force multiplier.
3935
+
3936
+ 00:52:48.715 --> 00:52:51.328
3937
+ So for every I want every hour of
3938
+
3939
+ 00:52:51.328 --> 00:52:53.039
3940
+ effort that you put into the course to
3941
+
3942
+ 00:52:53.040 --> 00:52:55.150
3943
+ produce as much learning as possible.
3944
+
3945
+ 00:52:55.150 --> 00:52:58.090
3946
+ So it means that you do need to put an
3947
+
3948
+ 00:52:58.090 --> 00:53:00.419
3949
+ effort to get value out of the course,
3950
+
3951
+ 00:53:00.420 --> 00:53:03.040
3952
+ but I try to select the topics and
3953
+
3954
+ 00:53:03.040 --> 00:53:03.800
3955
+ Assignments.
3956
+
3957
+ 00:53:04.920 --> 00:53:07.420
3958
+ And organize the materials in a way
3959
+
3960
+ 00:53:07.420 --> 00:53:09.450
3961
+ that makes your Learning efficient.
3962
+
3963
+ 00:53:11.860 --> 00:53:13.710
3964
+ Alright, so now what do you do next?
3965
+
3966
+ 00:53:13.710 --> 00:53:16.120
3967
+ Bookmark the website so that you can go
3968
+
3969
+ 00:53:16.120 --> 00:53:17.600
3970
+ back to it easily?
3971
+
3972
+ 00:53:17.600 --> 00:53:19.160
3973
+ Sign up for CampusWire?
3974
+
3975
+ 00:53:19.880 --> 00:53:22.300
3976
+ Read the syllabus and the schedule and
3977
+
3978
+ 00:53:22.300 --> 00:53:24.330
3979
+ I would recommend watching the
3980
+
3981
+ 00:53:24.330 --> 00:53:25.000
3982
+ tutorials.
3983
+
3984
+ 00:53:25.760 --> 00:53:27.650
3985
+ And then in the next class I'm going to
3986
+
3987
+ 00:53:27.650 --> 00:53:30.570
3988
+ talk about K nearest neighbor and kind
3989
+
3990
+ 00:53:30.570 --> 00:53:32.740
3991
+ of an overview of classification and
3992
+
3993
+ 00:53:32.740 --> 00:53:33.710
3994
+ regression.
3995
+
3996
+ 00:53:33.710 --> 00:53:35.843
3997
+ And I'll also introduce homework 1 S
3998
+
3999
+ 00:53:35.843 --> 00:53:37.400
4000
+ you can check out homework one if you
4001
+
4002
+ 00:53:37.400 --> 00:53:40.059
4003
+ want, or just wait till the next class
4004
+
4005
+ 00:53:40.060 --> 00:53:41.110
4006
+ when I introduce it.
4007
+
4008
+ 00:53:42.990 --> 00:53:45.040
4009
+ Alright, so that's it for today.
4010
+
4011
+ 00:53:45.040 --> 00:53:48.420
4012
+ I'll first before you get up, does
4013
+
4014
+ 00:53:48.420 --> 00:53:50.000
4015
+ anybody have any questions that you
4016
+
4017
+ 00:53:50.000 --> 00:53:51.420
4018
+ want to ask?
4019
+
4020
+ 00:53:51.420 --> 00:53:54.460
4021
+ Wait, don't get up, don't pick up your
4022
+
4023
+ 00:53:54.460 --> 00:53:55.120
4024
+ things yet.
4025
+
4026
+ 00:53:55.120 --> 00:53:55.960
4027
+ Yes.
4028
+
4029
+ 00:53:56.240 --> 00:53:56.680
4030
+ Join the.
4031
+
4032
+ 00:53:58.240 --> 00:54:00.240
4033
+ OK.
4034
+
4035
+ 00:54:00.240 --> 00:54:01.000
4036
+ Anything else?
4037
+
4038
+ 00:54:01.780 --> 00:54:03.560
4039
+ Alright, so just come up to me if you
4040
+
4041
+ 00:54:03.560 --> 00:54:05.730
4042
+ have any questions and I'll be here for
4043
+
4044
+ 00:54:05.730 --> 00:54:06.070
4045
+ a while.
4046
+
4047
+ 00:54:07.630 --> 00:54:08.600
4048
+ See you Thursday.
4049
+
4050
+ 00:54:10.050 --> 00:54:12.390
4051
+ A limited submission on the exams.
4052
+
4053
+ 00:54:12.390 --> 00:54:15.270
4054
+ What do you have like?
4055
+
4056
+ 01:14:27.910 --> 01:14:29.830
4057
+ Testing, testing, testing.
4058
+
4059
+ 01:14:40.680 --> 01:14:41.020
4060
+ OK.
4061
+
4062
+ 01:14:43.690 --> 01:14:44.490
4063
+ Every time you come.
4064
+
CS_441_2023_Spring_January_19,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1444ebe5da7801f4b7d5035584ae22f3face9a63131c5bcd6eb6a21f2e075b18
3
+ size 76791576
CS_441_2023_Spring_January_19,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_January_24,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2421310e306d45f4b1bbcb14f85bc1386c64f99ae840e57cbb68e1b6bc4be765
3
+ size 76788888
CS_441_2023_Spring_January_24,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_January_26,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0d1f5e1f85c835f71ca297f3ca04d500118098d8a9b9dd270ea55ba6eff731a
3
+ size 76791576
CS_441_2023_Spring_January_26,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff
 
CS_441_2023_Spring_January_31,_2023.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67486e63b34e48954bfdd7d80580cadd9364b308a4e7f1d3840dbb0949f52f14
3
+ size 76791960
CS_441_2023_Spring_January_31,_2023.vtt ADDED
The diff for this file is too large to render. See raw diff