Mercurial > pylearn

comparison algorithms/logistic_regression.py @ 497:a272f4cbf004

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
'x' => 'input' 'y' => 'output'
author Joseph Turian <turian@gmail.com>
date Tue, 28 Oct 2008 12:25:04 -0400
parents 7560817a07e8
children a419edf4e06c
comparison
equal deleted inserted replaced
496:f13847478c6d 497:a272f4cbf004
16 self.w = N.zeros((n_in, n_out)) 16 self.w = N.zeros((n_in, n_out))
17 self.b = N.zeros(n_out) 17 self.b = N.zeros(n_out)
18 self.lr = 0.01 18 self.lr = 0.01
19 self.__hide__ = ['params'] 19 self.__hide__ = ['params']
20 20
21 def __init__(self, x=None, targ=None, w=None, b=None, lr=None, regularize=False): 21 def __init__(self, input=None, targ=None, w=None, b=None, lr=None, regularize=False):
22 super(Module_Nclass, self).__init__() #boilerplate 22 super(Module_Nclass, self).__init__() #boilerplate
23 23
24 self.x = x if x is not None else T.matrix() 24 self.input = input if input is not None else T.matrix('input')
25 self.targ = targ if targ is not None else T.lvector() 25 self.targ = targ if targ is not None else T.lvector()
26 26
27 self.w = w if w is not None else module.Member(T.dmatrix()) 27 self.w = w if w is not None else module.Member(T.dmatrix())
28 self.b = b if b is not None else module.Member(T.dvector()) 28 self.b = b if b is not None else module.Member(T.dvector())
29 self.lr = lr if lr is not None else module.Member(T.dscalar()) 29 self.lr = lr if lr is not None else module.Member(T.dscalar())
30 30
31 self.params = [p for p in [self.w, self.b] if p.owner is None] 31 self.params = [p for p in [self.w, self.b] if p.owner is None]
32 32
33 xent, y = nnet.crossentropy_softmax_1hot( 33 xent, output = nnet.crossentropy_softmax_1hot(
34 T.dot(self.x, self.w) + self.b, self.targ) 34 T.dot(self.input, self.w) + self.b, self.targ)
35 sum_xent = T.sum(xent) 35 sum_xent = T.sum(xent)
36 36
37 self.y = y 37 self.output = output
38 self.sum_xent = sum_xent 38 self.sum_xent = sum_xent
39 self.cost = sum_xent 39 self.cost = sum_xent
40 40
41 #define the apply method 41 #define the apply method
42 self.pred = T.argmax(T.dot(self.x, self.w) + self.b, axis=1) 42 self.pred = T.argmax(T.dot(self.input, self.w) + self.b, axis=1)
43 self.apply = module.Method([self.x], self.pred) 43 self.apply = module.Method([self.input], self.pred)
44 44
45 if self.params: 45 if self.params:
46 gparams = T.grad(sum_xent, self.params) 46 gparams = T.grad(sum_xent, self.params)
47 47
48 self.update = module.Method([self.x, self.targ], sum_xent, 48 self.update = module.Method([self.input, self.targ], sum_xent,
49 updates = dict((p, p - self.lr * g) for p, g in zip(self.params, gparams))) 49 updates = dict((p, p - self.lr * g) for p, g in zip(self.params, gparams)))
50 50
51 class Module(module.FancyModule): 51 class Module(module.FancyModule):
52 class InstanceType(module.FancyModuleInstance): 52 class InstanceType(module.FancyModuleInstance):
53 def initialize(self, n_in): 53 def initialize(self, n_in):
56 self.w = N.random.randn(n_in,1) 56 self.w = N.random.randn(n_in,1)
57 self.b = N.random.randn(1) 57 self.b = N.random.randn(1)
58 self.lr = 0.01 58 self.lr = 0.01
59 self.__hide__ = ['params'] 59 self.__hide__ = ['params']
60 60
61 def __init__(self, x=None, targ=None, w=None, b=None, lr=None, regularize=False): 61 def __init__(self, input=None, targ=None, w=None, b=None, lr=None, regularize=False):
62 super(Module, self).__init__() #boilerplate 62 super(Module, self).__init__() #boilerplate
63 63
64 self.x = x if x is not None else T.matrix() 64 self.input = input if input is not None else T.matrix('input')
65 self.targ = targ if targ is not None else T.lcol() 65 self.targ = targ if targ is not None else T.lcol()
66 66
67 self.w = w if w is not None else module.Member(T.dmatrix()) 67 self.w = w if w is not None else module.Member(T.dmatrix())
68 self.b = b if b is not None else module.Member(T.dvector()) 68 self.b = b if b is not None else module.Member(T.dvector())
69 self.lr = lr if lr is not None else module.Member(T.dscalar()) 69 self.lr = lr if lr is not None else module.Member(T.dscalar())
70 70
71 self.params = [p for p in [self.w, self.b] if p.owner is None] 71 self.params = [p for p in [self.w, self.b] if p.owner is None]
72 72
73 y = nnet.sigmoid(T.dot(self.x, self.w)) 73 output = nnet.sigmoid(T.dot(self.x, self.w))
74 xent = -self.targ * T.log(y) - (1.0 - self.targ) * T.log(1.0 - y) 74 xent = -self.targ * T.log(output) - (1.0 - self.targ) * T.log(1.0 - output)
75 sum_xent = T.sum(xent) 75 sum_xent = T.sum(xent)
76 76
77 self.y = y 77 self.output = output
78 self.xent = xent 78 self.xent = xent
79 self.sum_xent = sum_xent 79 self.sum_xent = sum_xent
80 self.cost = sum_xent 80 self.cost = sum_xent
81 81
82 #define the apply method 82 #define the apply method
83 self.pred = (T.dot(self.x, self.w) + self.b) > 0.0 83 self.pred = (T.dot(self.input, self.w) + self.b) > 0.0
84 self.apply = module.Method([self.x], self.pred) 84 self.apply = module.Method([self.input], self.pred)
85 85
86 #if this module has any internal parameters, define an update function for them 86 #if this module has any internal parameters, define an update function for them
87 if self.params: 87 if self.params:
88 gparams = T.grad(sum_xent, self.params) 88 gparams = T.grad(sum_xent, self.params)
89 self.update = module.Method([self.x, self.targ], sum_xent, 89 self.update = module.Method([self.input, self.targ], sum_xent,
90 updates = dict((p, p - self.lr * g) for p, g in zip(self.params, gparams))) 90 updates = dict((p, p - self.lr * g) for p, g in zip(self.params, gparams)))
91 91
92 92
93 93
94 class Learner(object): 94 class Learner(object):