Mercurial > pylearn
comparison doc/v2_planning/main_plan.txt @ 1189:0e12ea6ba661
fix many rst syntax error warning.
author | Frederic Bastien <nouiz@nouiz.org> |
---|---|
date | Fri, 17 Sep 2010 20:55:18 -0400 |
parents | 1ed0719cfbce |
children |
comparison
equal
deleted
inserted
replaced
1188:073c2fab7bcd | 1189:0e12ea6ba661 |
---|---|
1 | 1 |
2 Motivation | 2 Motivation |
3 ========== | 3 ========== |
4 | 4 |
5 Yoshua (points discussed Thursday Sept 2, 2010 at LISA tea-talk) | 5 Yoshua (points discussed Thursday Sept 2, 2010 at LISA tea-talk) |
6 ------ | 6 ---------------------------------------------------------------- |
7 | 7 |
8 ****** Why we need to get better organized in our code-writing ****** | 8 ****** Why we need to get better organized in our code-writing ****** |
9 | 9 |
10 - current state of affairs on top of Theano is anarchic and does not lend itself to easy code re-use | 10 - current state of affairs on top of Theano is anarchic and does not lend itself to easy code re-use |
11 - the lab is growing and will continue to grow significantly, and more people outside the lab are using Theano | 11 - the lab is growing and will continue to grow significantly, and more people outside the lab are using Theano |
149 | 149 |
150 | 150 |
151 Another thing to consider related to datasets is that there are a number of | 151 Another thing to consider related to datasets is that there are a number of |
152 other efforts to have standard ML datasets, and we should be aware of them, | 152 other efforts to have standard ML datasets, and we should be aware of them, |
153 and compatible with them when it's easy: | 153 and compatible with them when it's easy: |
154 | |
154 - mldata.org (they have a file format, not sure how many use it) | 155 - mldata.org (they have a file format, not sure how many use it) |
155 - weka (ARFF file format) | 156 - weka (ARFF file format) |
156 - scikits.learn | 157 - scikits.learn |
157 - hdf5 / pytables | 158 - hdf5 / pytables |
158 | 159 |
166 found in pylearn. | 167 found in pylearn. |
167 | 168 |
168 Yoshua (about ideas proposed by Pascal Vincent a while ago): | 169 Yoshua (about ideas proposed by Pascal Vincent a while ago): |
169 | 170 |
170 - we may want to distinguish between datasets and tasks: a task defines | 171 - we may want to distinguish between datasets and tasks: a task defines |
171 not just the data but also things like what is the input and what is the | 172 not just the data but also things like what is the input and what is the |
172 target (for supervised learning), and *importantly* a set of performance metrics | 173 target (for supervised learning), and *importantly* a set of performance metrics |
173 that make sense for this task (e.g. those used by papers solving a particular | 174 that make sense for this task (e.g. those used by papers solving a particular |
174 task, or reported for a particular benchmark) | 175 task, or reported for a particular benchmark) |
175 | 176 |
176 - we should discuss about a few "standards" that datasets and tasks may comply to, such as | 177 - we should discuss about a few "standards" that datasets and tasks may comply to, such as |
177 - "input" and "target" fields inside each example, for supervised or semi-supervised learning tasks | 178 - "input" and "target" fields inside each example, for supervised or semi-supervised learning tasks |
178 (with a convention for the semi-supervised case when only the input or only the target is observed) | 179 (with a convention for the semi-supervised case when only the input or only the target is observed) |
179 - "input" for unsupervised learning | 180 - "input" for unsupervised learning |