Mercurial > pylearn
comparison doc/v2_planning/dataset.txt @ 1054:a474fabd1f37
v2_planning dataset - added questions
| author | James Bergstra <bergstrj@iro.umontreal.ca> |
|---|---|
| date | Wed, 08 Sep 2010 16:20:20 -0400 |
| parents | 1b61cbe0810b |
| children | 20a1af112a75 |
comparison
equal
deleted
inserted
replaced
| 1053:390166ace9e5 | 1054:a474fabd1f37 |
|---|---|
| 151 batches_valid = range(int(mnist_data.n_batches*0.8),mnist_data.n_batches) | 151 batches_valid = range(int(mnist_data.n_batches*0.8),mnist_data.n_batches) |
| 152 | 152 |
| 153 xt,yt = mnist_data.get_batch(batches_train[0]) | 153 xt,yt = mnist_data.get_batch(batches_train[0]) |
| 154 xv,yv = mnist_data.get_batch(batches_valid[0]) | 154 xv,yv = mnist_data.get_batch(batches_valid[0]) |
| 155 | 155 |
| 156 | |
| 157 | |
| 158 | |
| 159 COMMENTS | |
| 160 ~~~~~~~~ | |
| 161 | |
| 162 | |
| 163 JB asks: What may be passed as argument to the functions in Dataset, and what | |
| 164 can be expected in return? Are there side effects (e.g. on the state of the | |
| 165 Dataset) associated with any of the functions? | |
| 166 | |
| 167 JB asks: What properties are part of the Dataset API? What possible types can | |
| 168 they have, are they expected to be read-only or writeable? What do they mean? | |
| 169 | |
| 170 | |
| 171 JB asks: What is a view? Does set_view change the Dataset or return a new | |
| 172 Dataset with a certain view of the original (in which case call it get_view)? | |
| 173 Does the view imply the types of the return-value of functions like | |
| 174 get_batch? What is the difference between the view and the subclasses of | |
| 175 Dataset in PyML? | |
| 176 | |
| 177 JB asks: Do container formats (I'm thinking of HDF5) offer features for fast | |
| 178 retrieval that we would like to expose via this interface? | |
| 179 | |
| 180 JB asks: How would you recommend using this sort of dataset in a boosting | |
| 181 algorithm where points need to be re-weighted. | |
| 182 | |
| 183 | |
| 184 JB asks: Do we want to provide for the possibility of feedback that modifies the | |
| 185 dataset? For example, curriculum learning might be adaptive in this sense, or | |
| 186 if we wanted to provide a virtual world for an agent as a dataset then we need | |
| 187 to provide 'actions' to get the next batch. Could this be done in the current | |
| 188 API? | |
| 189 | |
| 190 | |
| 191 |