Mercurial > pylearn

comparison linear_regression.py @ 427:fa4a5fee53ce

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Showing the path to an online version of linear regressor.
author Yoshua Bengio <bengioy@iro.umontreal.ca>
date Tue, 22 Jul 2008 16:44:42 -0400
parents e01f17be270a
children 0f8c81b0776d
comparison
equal deleted inserted replaced
426:d7611a3811f2 427:fa4a5fee53ce
33 The test_set must have field "input", and needs "target" if 33 The test_set must have field "input", and needs "target" if
34 we want to compute the squared errors. 34 we want to compute the squared errors.
35 35
36 The predictor parameters are obtained analytically from the training set. 36 The predictor parameters are obtained analytically from the training set.
37 37
38 *** NOT IMPLEMENTED YET ***
39 Training can proceed sequentially (with multiple calls to update with
40 different disjoint subsets of the training sets). After each call to
41 update the predictor is ready to be used (and optimized for the union
42 of all the training sets passed to update since construction or since
43 the last call to forget).
44 ***************************
45
46 For each (input[t],output[t]) pair in a minibatch,:: 38 For each (input[t],output[t]) pair in a minibatch,::
47 39
48 output_t = b + W * input_t 40 output_t = b + W * input_t
49 41
50 where b and W are obtained by minimizing:: 42 where b and W are obtained by minimizing::
187 return CachedDataSet(ds) 179 return CachedDataSet(ds)
188 else: 180 else:
189 return ds 181 return ds
190 182
191 183
184 #TODO : an online version
185 class OnlineLinearRegression(OnlineLearningAlgorithm):
186 """
187 Training can proceed sequentially (with multiple calls to update with
188 different disjoint subsets of the training sets). After each call to
189 update the predictor is ready to be used (and optimized for the union
190 of all the training sets passed to update since construction or since
191 the last call to forget).
192 """
193 pass
194
195
196
197