|
|
d369abc61b
|
Support retrieving nn gradient wrt inputs in backpropagation & add tests
|
2022-01-23 19:37:42 +03:00 |
|
|
|
7040faecd4
|
Add identity activation function to ml
|
2022-01-23 19:36:38 +03:00 |
|
|
|
3dbbcadfe3
|
Add generic backpropagation (supporting any loss function, not only l2)
|
2022-01-21 18:42:48 +03:00 |
|
|
|
8630525dcf
|
Add ml::neural_net_learner implementation with backpropagation, gradient descent, and tests
|
2022-01-21 12:12:38 +03:00 |
|
|
|
def0615baa
|
Add ml::activation_type::count
|
2022-01-21 12:12:08 +03:00 |
|
|
|
d410c970e6
|
Add an utility function to compute l2 loss in ml
|
2022-01-21 12:11:38 +03:00 |
|
|
|
6d40e4dd98
|
Add some empty lines
|
2022-01-21 12:11:14 +03:00 |
|
|
|
b47b6aa224
|
Fix ml::unknown_activation_type overriding what()
|
2022-01-21 12:10:51 +03:00 |
|
|
|
11b2d002ba
|
Move neural net errors to a separate file
|
2022-01-20 20:07:36 +03:00 |
|
|
|
2f1afc368e
|
Rename evaluate -> apply
|
2022-01-20 14:19:10 +03:00 |
|
|
|
3a3fffa232
|
Add neural net random initialization
|
2022-01-20 14:16:54 +03:00 |
|
|
|
587a80b0de
|
Move neural net stuff to a separate subdirectory of ml
|
2022-01-20 14:16:22 +03:00 |
|
|
|
d37edd207b
|
Make ml::neural_net return spans
|
2022-01-20 13:52:13 +03:00 |
|
|
|
0ab0ce9134
|
Remove unused include
|
2022-01-20 13:48:25 +03:00 |
|
|
|
434c327a9d
|
Throw an exception if number of neural net layer sizes & activation types do not match
|
2022-01-20 13:46:09 +03:00 |
|
|
|
b2fea97c8f
|
Add new library for ML stuff & a neural net implementation
|
2022-01-19 20:12:15 +03:00 |
|
