Mercurial > pylearn

comparison sparse_random_autoassociator/graph.py @ 383:344d1f874af7

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Small fix
author Joseph Turian <turian@gmail.com>
date Tue, 08 Jul 2008 01:59:42 -0400
parents b4efd192d880
children edec18614a70
comparison
equal deleted inserted replaced
382:b4efd192d880 383:344d1f874af7
4 @todo: Make nearly everything private. 4 @todo: Make nearly everything private.
5 """ 5 """
6 6
7 from globals import MARGIN 7 from globals import MARGIN
8 8
9 from pylearn.nnet_ops import sigmoid, crossentropy_softmax_1hot 9 from pylearn.nnet_ops import sigmoid, binary_crossentropy
10 from theano import tensor as t 10 from theano import tensor as t
11 from theano.tensor import dot 11 from theano.tensor import dot
12 xnonzero = t.dvector() 12 xnonzero = t.dvector()
13 w1nonzero = t.dmatrix() 13 w1nonzero = t.dmatrix()
14 b1 = t.dvector() 14 b1 = t.dvector()
27 nonzeroloss = hingeloss(ynonzero - t.max(yzero) - MARGIN) 27 nonzeroloss = hingeloss(ynonzero - t.max(yzero) - MARGIN)
28 zeroloss = hingeloss(-t.max(-(ynonzero)) - yzero - MARGIN) 28 zeroloss = hingeloss(-t.max(-(ynonzero)) - yzero - MARGIN)
29 # xnonzero sensitive loss: 29 # xnonzero sensitive loss:
30 #nonzeroloss = hingeloss(ynonzero - t.max(yzero) - MARGIN - xnonzero) 30 #nonzeroloss = hingeloss(ynonzero - t.max(yzero) - MARGIN - xnonzero)
31 #zeroloss = hingeloss(-t.max(-(ynonzero - xnonzero)) - yzero - MARGIN) 31 #zeroloss = hingeloss(-t.max(-(ynonzero - xnonzero)) - yzero - MARGIN)
32 loss = t.sum(nonzeroloss) + t.sum(zeroloss) 32 #loss = t.sum(nonzeroloss) + t.sum(zeroloss)
33 33
34 #loss = t.sum(binary_crossentropy(ynonzero, xnonzero)) + t.sum(binary_crossentropy(yzero, t.constant(0))) 34 loss = t.sum(binary_crossentropy(ynonzero, xnonzero)) + t.sum(binary_crossentropy(yzero, t.constant(0)))
35 35
36 (gw1nonzero, gb1, gw2nonzero, gw2zero, gb2nonzero, gb2zero) = t.grad(loss, [w1nonzero, b1, w2nonzero, w2zero, b2nonzero, b2zero]) 36 (gw1nonzero, gb1, gw2nonzero, gw2zero, gb2nonzero, gb2zero) = t.grad(loss, [w1nonzero, b1, w2nonzero, w2zero, b2nonzero, b2zero])
37 37
38 import theano.compile 38 import theano.compile
39 39