Mercurial > pylearn

comparison mlp_factory_approach.py @ 187:ebbb0e749565

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
added mlp_factory_approach
author James Bergstra <bergstrj@iro.umontreal.ca>
date Wed, 14 May 2008 11:51:08 -0400
parents
children 8f58abb943d4 f2ddc795ec49
comparison
equal deleted inserted replaced
186:562f308873f0 187:ebbb0e749565
1 import dataset
2 import theano
3 import theano.tensor as t
4 import numpy
5 import nnet_ops
6
7 def _randshape(*shape):
8 return (numpy.random.rand(*shape) -0.5) * 0.001
9 def _function(inputs, outputs, linker='c&py'):
10 return theano.function(inputs, outputs, unpack_single=False,linker=linker)
11
12 class NeuralNet(object):
13
14 class Model(object):
15 def __init__(self, nnet, params):
16 self.nnet = nnet
17 self.params = params
18
19 def update(self, trainset, stopper=None):
20 """Update this model from more training data."""
21 v = self.nnet.v
22 params = self.params
23 update_fn = _function([v.input, v.target] + v.params, [v.nll] + v.new_params)
24 if stopper is not None:
25 raise NotImplementedError()
26 else:
27 for i in xrange(100):
28 for input, target in trainset.minibatches(['input', 'target'],
29 minibatch_size=min(32, len(trainset))):
30 dummy = update_fn(input, target[:,0], *params)
31 if 0: print dummy[0] #the nll
32
33 def __call__(self, testset,
34 output_fieldnames=['output_class'],
35 test_stats_collector=None,
36 copy_inputs=False,
37 put_stats_in_output_dataset=True,
38 output_attributes=[]):
39 """Apply this model (as a function) to new data"""
40 inputs = [self.nnet.v.input, self.nnet.v.target] + self.nnet.v.params
41 fn = _function(inputs, [getattr(self.nnet.v, name) for name in output_fieldnames])
42 if 'target' in testset.fields():
43 return dataset.ApplyFunctionDataSet(testset,
44 lambda input, target: fn(input, target[:,0], *self.params),
45 output_fieldnames)
46 else:
47 return dataset.ApplyFunctionDataSet(testset,
48 lambda input: fn(input, numpy.zeros(1,dtype='int64'), *self.params),
49 output_fieldnames)
50
51 def __init__(self, ninputs, nhid, nclass, lr, nepochs,
52 l2coef=0.0,
53 linker='c&yp',
54 hidden_layer=None):
55 class Vars:
56 def __init__(self, lr, l2coef):
57 lr = t.constant(lr)
58 l2coef = t.constant(l2coef)
59 input = t.matrix('input') # n_examples x n_inputs
60 target = t.ivector('target') # n_examples x 1
61 W2 = t.matrix('W2')
62 b2 = t.vector('b2')
63
64 if hidden_layer:
65 hid, hid_params, hid_ivals, hid_regularization = hidden_layer(input)
66 else:
67 W1 = t.matrix('W1')
68 b1 = t.vector('b1')
69 hid = t.tanh(b1 + t.dot(input, W1))
70 hid_params = [W1, b1]
71 hid_regularization = l2coef * t.sum(W1*W1)
72 hid_ivals = lambda : [_randshape(ninputs, nhid), _randshape(nhid)]
73
74 params = [W2, b2] + hid_params
75 nll, predictions = nnet_ops.crossentropy_softmax_1hot( b2 + t.dot(hid, W2), target)
76 regularization = l2coef * t.sum(W2*W2) + hid_regularization
77 output_class = t.argmax(predictions,1)
78 loss_01 = t.neq(output_class, target)
79 g_params = t.grad(nll + regularization, params)
80 new_params = [t.sub_inplace(p, lr * gp) for p,gp in zip(params, g_params)]
81 self.__dict__.update(locals()); del self.self
82 self.nhid = nhid
83 self.nclass = nclass
84 self.nepochs = nepochs
85 self.v = Vars(lr, l2coef)
86 self.params = None
87
88 def __call__(self, trainset=None, iparams=None):
89 if iparams is None:
90 iparams = [_randshape(self.nhid, self.nclass), _randshape(self.nclass)]\
91 + self.v.hid_ivals()
92 rval = NeuralNet.Model(self, iparams)
93 if trainset:
94 rval.update(trainset)
95 return rval
96
97
98 if __name__ == '__main__':
99 training_set1 = dataset.ArrayDataSet(numpy.array([[0, 0, 0],
100 [0, 1, 1],
101 [1, 0, 1],
102 [1, 1, 1]]),
103 {'input':slice(2),'target':2})
104 training_set2 = dataset.ArrayDataSet(numpy.array([[0, 0, 0],
105 [0, 1, 1],
106 [1, 0, 0],
107 [1, 1, 1]]),
108 {'input':slice(2),'target':2})
109 test_data = dataset.ArrayDataSet(numpy.array([[0, 0, 0],
110 [0, 1, 1],
111 [1, 0, 0],
112 [1, 1, 1]]),
113 {'input':slice(2)})
114
115 learn_algo = NeuralNet(2, 10, 3, .1, 1000)
116
117 model1 = learn_algo(training_set1)
118
119 model2 = learn_algo(training_set2)
120
121 n_match = 0
122 for o1, o2 in zip(model1(test_data), model2(test_data)):
123 n_match += (o1 == o2)
124
125 print n_match, numpy.sum(training_set1.fields()['target'] ==
126 training_set2.fields()['target'])
127