Mercurial > pylearn
comparison simple_autoassociator.py/graph.py @ 386:a474341861fa
Added a simple AA
| author | Joseph Turian <turian@gmail.com> |
|---|---|
| date | Tue, 08 Jul 2008 02:27:00 -0400 |
| parents | sparse_random_autoassociator/graph.py@edec18614a70 |
| children | 98ca97cc9910 |
comparison
equal
deleted
inserted
replaced
| 385:db28ff3fb887 | 386:a474341861fa |
|---|---|
| 1 """ | |
| 2 Theano graph for a simple autoassociator. | |
| 3 @todo: Make nearly everything private. | |
| 4 """ | |
| 5 | |
| 6 from pylearn.nnet_ops import sigmoid, binary_crossentropy | |
| 7 from theano import tensor as t | |
| 8 from theano.tensor import dot | |
| 9 x = t.dvector() | |
| 10 w1 = t.dmatrix() | |
| 11 b1 = t.dvector() | |
| 12 w2 = t.dmatrix() | |
| 13 b2 = t.dvector() | |
| 14 h = sigmoid(dot(x, w1) + b1) | |
| 15 y = sigmoid(dot(h, w2) + b2) | |
| 16 | |
| 17 loss = t.sum(binary_crossentropy(y, x)) | |
| 18 | |
| 19 (gw1, gb1, gw2, gb2) = t.grad(loss, [w1, b1, w2, b2]) | |
| 20 | |
| 21 import theano.compile | |
| 22 | |
| 23 inputs = [x, w1, b1, w2, b2] | |
| 24 outputs = [y, loss, gw1, gb1, gw2, gb2] | |
| 25 trainfn = theano.compile.function(inputs, outputs) |