comparison transformations/pipeline.py @ 116:3bec123dd75d

changes on pipeline mecanism: we now sample a different complexity for each transformations, this because when we use the same sampled complexity for all the modules 1/8 of the time we are close to 0 and we obtain an image very close to the source, we now save a complexity for each module in the parameters array
author Xavier Glorot <glorotxa@iro.umontreal.ca>
date Wed, 17 Feb 2010 16:22:54 -0500
parents b84a0d009af8
children ccce06590e64
comparison
equal deleted inserted replaced
115:b84a0d009af8 116:3bec123dd75d
118 118
119 img = img.reshape(img_size) 119 img = img.reshape(img_size)
120 120
121 param_idx = 0 121 param_idx = 0
122 mod_idx = 0 122 mod_idx = 0
123 # store complexity along with other params
124 self.params[global_idx, 0] = complexity
125 for mod in self.modules: 123 for mod in self.modules:
126 # This used to be done _per batch_, 124 # This used to be done _per batch_,
127 # ie. out of the "for img" loop 125 # ie. out of the "for img" loop
128 complexity = complexity_iterator.next() 126 complexity = complexity_iterator.next()
129 #better to do a complexity sampling for each transformations in order to have more variability 127 #better to do a complexity sampling for each transformations in order to have more variability