DOC Fix incorrect 0-1 scaling in the RBM example #19363
Conversation
|
Thank you for taking a look! I've updated the MR to use the Are the related docs rebuilded automatically when deployed or should I update anything else? BTW, proper scaling changes the results slightly (generally for the better if I understand correctly). Before/after diff: Logistic regression using RBM features:
precision recall f1-score support
0 0.99 0.98 0.99 174
- 1 0.91 0.93 0.92 184
+ 1 0.92 0.93 0.93 184
2 0.92 0.96 0.94 166
- 3 0.93 0.88 0.90 194
+ 3 0.94 0.88 0.91 194
4 0.97 0.94 0.95 186
- 5 0.94 0.91 0.92 181
- 6 0.97 0.96 0.97 207
+ 5 0.93 0.92 0.92 181
+ 6 0.97 0.97 0.97 207
7 0.94 0.99 0.97 154
- 8 0.89 0.90 0.89 182
- 9 0.87 0.90 0.89 169
+ 8 0.89 0.89 0.89 182
+ 9 0.88 0.91 0.89 169
- accuracy 0.93 1797
- macro avg 0.93 0.94 0.93 1797
-weighted avg 0.93 0.93 0.93 1797
+ accuracy 0.94 1797
+ macro avg 0.94 0.94 0.94 1797
+weighted avg 0.94 0.94 0.94 1797
Logistic regression using raw pixel features:
precision recall f1-score support
0 0.90 0.92 0.91 174
- 1 0.60 0.59 0.60 184
- 2 0.75 0.85 0.80 166
- 3 0.77 0.78 0.78 194
- 4 0.82 0.84 0.83 186
+ 1 0.60 0.58 0.59 184
+ 2 0.76 0.85 0.80 166
+ 3 0.78 0.78 0.78 194
+ 4 0.81 0.84 0.83 186
5 0.77 0.76 0.76 181
6 0.90 0.87 0.89 207
7 0.85 0.88 0.87 154
8 0.67 0.58 0.62 182
9 0.74 0.76 0.75 169
accuracy 0.78 1797
macro avg 0.78 0.78 0.78 1797
weighted avg 0.78 0.78 0.78 1797
|
You don't have to update anything: the Edit: see this part of the configuration for the generation of Lines 326 to 346 in 468c3f4 |
|
Thanks for the review and providing relevant configuration snippet! Now it totally makes sense. |
Fix incorrect 0-1 scaling method in the "Restricted Boltzmann Machine features for digit classification" example.
Therefore, fix #19362.