Mercurial > pylearn

comparison doc/v2_planning/api_optimization.txt @ 1102:e7c52923f122

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
api_optimization: Couple replies
author Olivier Delalleau <delallea@iro>
date Mon, 13 Sep 2010 15:03:40 -0400
parents 153cf820a975
children c5c7ba805a2f
comparison
equal deleted inserted replaced
1101:b422cbaddc52 1102:e7c52923f122
112 OD asks: Could it be more convenient for x0 to be a list? 112 OD asks: Could it be more convenient for x0 to be a list?
113 113
114 JB replies: Yes, but that's not the interface used by other minimize() 114 JB replies: Yes, but that's not the interface used by other minimize()
115 routines (e.g. in scipy). Maybe another list-based interface is required? 115 routines (e.g. in scipy). Maybe another list-based interface is required?
116 116
117 OD replies: I think most people would prefer to use a list-based interface, so
118 they don't have to manually pack / unpack multiple arrrays of parameters. So I
119 would vote in favor or having both (where the main reason to also provide a
120 non-list interface would be to allow one to easily switch e.g. to scipy's
121 minimize).
122 I would guess the reason scipy's interface is like this is because it makes
123 it easier for the optimization algorithm, and the packing / unpacking is
124 typically done only once for non-iterative algorithms. However, the first
125 point does not matter if we are just wrapping a theano-based algorithm that
126 already has to handle multiple parameters (and the second may not hold if
127 we allow iterative algorithms).
117 128
118 OD asks: Why make a difference between iterative and one-shot versions? A one-shot 129 OD asks: Why make a difference between iterative and one-shot versions? A one-shot
119 algorithm can be seen as an iterative one that stops after its first 130 algorithm can be seen as an iterative one that stops after its first
120 iteration. The difference I see between the two interfaces proposed here 131 iteration. The difference I see between the two interfaces proposed here
121 is mostly that one relies on Theano while the other one does not, but 132 is mostly that one relies on Theano while the other one does not, but
125 JB replies: Right, it would make more sense to distinguish them by the fact that 136 JB replies: Right, it would make more sense to distinguish them by the fact that
126 one works on Theano objects, and the other on general Python callable functions. 137 one works on Theano objects, and the other on general Python callable functions.
127 There is room for an iterative numpy interface, but I didn't make it yet. Would 138 There is room for an iterative numpy interface, but I didn't make it yet. Would
128 that answer your question? 139 that answer your question?
129 140
130 141 OD replies and asks: Partly. Do we really need a non-iterative interface?
131 142
132 Examples 143 Examples
133 -------- 144 --------
134 145
135 146