Mercurial > pylearn

comparison dataset.py @ 216:4b7e89b75e2b

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Modified ArrayDataSet's handling of column fields. Previously, if a fieldname were associated with an integer column index (by opposition to a column range or slice) then it would be returned as a Nx1 matrix. Now if a fieldname is associated with an integer column index, then it will make a field which is a vector of length N. The old behaviour can still be achieved by associating a fieldname with the slice(col, col+1).
author James Bergstra <bergstrj@iro.umontreal.ca>
date Thu, 22 May 2008 19:07:51 -0400
parents bd728c83faff
children df3fae88ab46
comparison
equal deleted inserted replaced
215:6fa8fbb0c3f6 216:4b7e89b75e2b
243 assert offset+minibatch_size<=self.L 243 assert offset+minibatch_size<=self.L
244 ds_nbatches = (self.L-self.next_row)/self.minibatch_size 244 ds_nbatches = (self.L-self.next_row)/self.minibatch_size
245 if n_batches is not None: 245 if n_batches is not None:
246 ds_nbatches = min(n_batches,ds_nbatches) 246 ds_nbatches = min(n_batches,ds_nbatches)
247 if fieldnames: 247 if fieldnames:
248 assert dataset.hasFields(*fieldnames) 248 if not dataset.hasFields(*fieldnames):
249 raise ValueError('field not present', fieldnames)
249 else: 250 else:
250 self.fieldnames=dataset.fieldNames() 251 self.fieldnames=dataset.fieldNames()
251 self.iterator = self.dataset.minibatches_nowrap(self.fieldnames,self.minibatch_size, 252 self.iterator = self.dataset.minibatches_nowrap(self.fieldnames,self.minibatch_size,
252 ds_nbatches,self.next_row) 253 ds_nbatches,self.next_row)
253 254
967 968
968 # check consistency and complete slices definitions 969 # check consistency and complete slices definitions
969 for fieldname, fieldcolumns in self.fields_columns.items(): 970 for fieldname, fieldcolumns in self.fields_columns.items():
970 if type(fieldcolumns) is int: 971 if type(fieldcolumns) is int:
971 assert fieldcolumns>=0 and fieldcolumns<data_array.shape[1] 972 assert fieldcolumns>=0 and fieldcolumns<data_array.shape[1]
972 self.fields_columns[fieldname]=[fieldcolumns] 973
974 if 0:
975 #I changed this because it didn't make sense to me,
976 # and it made it more difficult to write my learner.
977 # If it breaks stuff, let's talk about it.
978 # - James 22/05/2008
979 self.fields_columns[fieldname]=[fieldcolumns]
980 else:
981 self.fields_columns[fieldname]=fieldcolumns
982
973 elif type(fieldcolumns) is slice: 983 elif type(fieldcolumns) is slice:
974 start,step=None,None 984 start,step=None,None
975 if not fieldcolumns.start: 985 if not fieldcolumns.start:
976 start=0 986 start=0
977 if not fieldcolumns.step: 987 if not fieldcolumns.step: