Joe Chi
commited on
Commit
•
697a68e
1
Parent(s):
8270912
Add gradio app
Browse files
app.py
ADDED
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from sklearn.decomposition import PCA
|
2 |
+
import matplotlib
|
3 |
+
matplotlib.use("Agg")
|
4 |
+
import matplotlib.pyplot as plt
|
5 |
+
import numpy as np
|
6 |
+
from scipy import stats
|
7 |
+
import gradio as gr
|
8 |
+
|
9 |
+
e = np.exp(1)
|
10 |
+
np.random.seed(4)
|
11 |
+
|
12 |
+
|
13 |
+
def pdf(x):
|
14 |
+
return 0.5 * (stats.norm(scale=0.25 / e).pdf(x) + stats.norm(scale=4 / e).pdf(x))
|
15 |
+
|
16 |
+
|
17 |
+
y = np.random.normal(scale=0.5, size=(30000))
|
18 |
+
x = np.random.normal(scale=0.5, size=(30000))
|
19 |
+
z = np.random.normal(scale=0.1, size=len(x))
|
20 |
+
|
21 |
+
density = pdf(x) * pdf(y)
|
22 |
+
pdf_z = pdf(5 * z)
|
23 |
+
|
24 |
+
density *= pdf_z
|
25 |
+
|
26 |
+
a = x + y
|
27 |
+
b = 2 * y
|
28 |
+
c = a - b + z
|
29 |
+
|
30 |
+
norm = np.sqrt(a.var() + b.var())
|
31 |
+
a /= norm
|
32 |
+
b /= norm
|
33 |
+
|
34 |
+
def plot_figs(fig_num, elev, azim):
|
35 |
+
fig = plt.figure()
|
36 |
+
plt.clf()
|
37 |
+
ax = fig.add_subplot(111, projection="3d", elev=elev, azim=azim)
|
38 |
+
ax.set_position([0, 0, 0.95, 1])
|
39 |
+
|
40 |
+
ax.scatter(a[::10], b[::10], c[::10], c=density[::10], marker="+", alpha=0.4)
|
41 |
+
Y = np.c_[a, b, c]
|
42 |
+
|
43 |
+
# Using SciPy's SVD, this would be:
|
44 |
+
# _, pca_score, Vt = scipy.linalg.svd(Y, full_matrices=False)
|
45 |
+
|
46 |
+
pca = PCA(n_components=3)
|
47 |
+
pca.fit(Y)
|
48 |
+
V = pca.components_.T
|
49 |
+
|
50 |
+
x_pca_axis, y_pca_axis, z_pca_axis = 3 * V
|
51 |
+
x_pca_plane = np.r_[x_pca_axis[:2], -x_pca_axis[1::-1]]
|
52 |
+
y_pca_plane = np.r_[y_pca_axis[:2], -y_pca_axis[1::-1]]
|
53 |
+
z_pca_plane = np.r_[z_pca_axis[:2], -z_pca_axis[1::-1]]
|
54 |
+
x_pca_plane.shape = (2, 2)
|
55 |
+
y_pca_plane.shape = (2, 2)
|
56 |
+
z_pca_plane.shape = (2, 2)
|
57 |
+
ax.plot_surface(x_pca_plane, y_pca_plane, z_pca_plane)
|
58 |
+
ax.xaxis.set_ticklabels([])
|
59 |
+
ax.yaxis.set_ticklabels([])
|
60 |
+
ax.zaxis.set_ticklabels([])
|
61 |
+
|
62 |
+
plt.savefig(f"{fig_num}.png")
|
63 |
+
|
64 |
+
return fig
|
65 |
+
|
66 |
+
def make_plot(plot_type):
|
67 |
+
if plot_type == "Very flat direction":
|
68 |
+
elev = -40
|
69 |
+
azim = -80
|
70 |
+
fig_num = 1
|
71 |
+
else:
|
72 |
+
elev = 30
|
73 |
+
azim = 20
|
74 |
+
fig_num = 2
|
75 |
+
plot_figs(fig_num, elev, azim)
|
76 |
+
|
77 |
+
title = "Principal components analysis (PCA)"
|
78 |
+
with gr.Blocks(title=title) as demo:
|
79 |
+
gr.Markdown(f"## {title}")
|
80 |
+
gr.Markdown("These figures aid in illustrating how a point cloud can be \
|
81 |
+
very flat in one direction–which is where PCA comes in to choose a direction that is not flat.")
|
82 |
+
|
83 |
+
|
84 |
+
button = gr.Radio(label="Plot type",
|
85 |
+
choices=['Very flat direction', 'Not flat direction'], value='Very flat direction')
|
86 |
+
plot = gr.Plot(label="Plot")
|
87 |
+
button.change(make_plot, inputs=button, outputs=[plot])
|
88 |
+
demo.load(make_plot, inputs=[button], outputs=[plot])
|
89 |
+
|
90 |
+
demo.launch()
|