Victor commited on
Commit
e4ed58f
1 Parent(s): 9a84cbe

Add description for isnnn_400.seed01 dataset.

Browse files
Files changed (1) hide show
  1. README.md +11 -2
README.md CHANGED
@@ -15,6 +15,11 @@ configs:
15
  path: isnnn_400/seed_01/d_train.parquet
16
  - split: test
17
  path: isnnn_400/seed_01/d_test.parquet
 
 
 
 
 
18
  - config_name: isnnn_400.seed01.gt_fim
19
  data_files:
20
  - split: test
@@ -111,15 +116,19 @@ configs:
111
  path: isnnn_400/seed_10/gt_fim.parquet
112
  ---
113
 
114
- # FIM-Estimation
 
115
  In a FIM-Estimation task, the input consists of the following components:
116
 
117
  1. A dataset of the form $\mathcal{D}_ {train} = \{(\mathbf{\lambda}_ i, x_ i)\}_ {i=1}^{\left|\mathcal{D}\right|}$, where $\vlambda_ i$ is a point in the parameter space and $x_ i$ is a sample.
118
  2. The structure of the samples $x_ i$. For example, if $x_ i\in\{0,1\}^n$ is a bitstring representing a measurement of a quantum system on a lattice with $n$ sites, then the said structure is the lattice and the correspondence between the bitstring bits and the lattice sites.
119
  3. Optionally, additional information about the statistical manifold. This could involve, e.g., the symmetries of the system used to generate the statistical manifold.
120
 
121
- The task is to estimate Fisher Information Metric as a matrix-valued function of parameter $\mathbf{\lambda}$.
122
 
 
 
 
123
  We (TODO:plan to) present datasets corresponding to 6 statistical manifolds: `XXZ300_Z`, `FIL24`, etc.
124
  The data is stored as follows:
125
  `data/xxz300_z` contains the data for `XXZ300_Z` datasets. Within that directory, there are subdirectories of the form `seed_??` each containing a dataset describing the same statistical manifold, but generated using a different seed (e.g. `seed_05`). We have `seed_05/d_train.parquet`, which is the dataset as described in #1 above. `seed_05/d_test.parquet` is a hold-out dataset which should not be used in estimation of $\mathbf{\lambda}$ (not even for hyperparameter tuning).
 
15
  path: isnnn_400/seed_01/d_train.parquet
16
  - split: test
17
  path: isnnn_400/seed_01/d_test.parquet
18
+ description: >-
19
+ Samples for IsNNN400 statistical manifold generated with seed=1.
20
+ `sample` column is formatted as bytearray of length 50 bytes (400 bits)
21
+ and represents a spin configuration of 20x20 lattice
22
+ with bit 0 corresponding to Z=+1 and bit 1 corresponding to Z=-1.
23
  - config_name: isnnn_400.seed01.gt_fim
24
  data_files:
25
  - split: test
 
116
  path: isnnn_400/seed_10/gt_fim.parquet
117
  ---
118
 
119
+ # Dataset Card for FIM-Estimation
120
+ ## Dataset Summary
121
  In a FIM-Estimation task, the input consists of the following components:
122
 
123
  1. A dataset of the form $\mathcal{D}_ {train} = \{(\mathbf{\lambda}_ i, x_ i)\}_ {i=1}^{\left|\mathcal{D}\right|}$, where $\vlambda_ i$ is a point in the parameter space and $x_ i$ is a sample.
124
  2. The structure of the samples $x_ i$. For example, if $x_ i\in\{0,1\}^n$ is a bitstring representing a measurement of a quantum system on a lattice with $n$ sites, then the said structure is the lattice and the correspondence between the bitstring bits and the lattice sites.
125
  3. Optionally, additional information about the statistical manifold. This could involve, e.g., the symmetries of the system used to generate the statistical manifold.
126
 
127
+ The task is to estimate Fisher Information Metric (FIM) as a matrix-valued function of parameter $\mathbf{\lambda}$.
128
 
129
+ In this dataset we present multiple such tasks (parameterised by task name and random seed). E.g. in order to solve the IsNNN400 task with seed=4 one would download config `isnnn_400.seed04`, split `train`, estimate the FIM, and compare the results with `isnnn_400.seed04.gt_fim`.
130
+
131
+ ## Additional information
132
  We (TODO:plan to) present datasets corresponding to 6 statistical manifolds: `XXZ300_Z`, `FIL24`, etc.
133
  The data is stored as follows:
134
  `data/xxz300_z` contains the data for `XXZ300_Z` datasets. Within that directory, there are subdirectories of the form `seed_??` each containing a dataset describing the same statistical manifold, but generated using a different seed (e.g. `seed_05`). We have `seed_05/d_train.parquet`, which is the dataset as described in #1 above. `seed_05/d_test.parquet` is a hold-out dataset which should not be used in estimation of $\mathbf{\lambda}$ (not even for hyperparameter tuning).