Mercurial > pylearn
comparison doc/v2_planning/datalearn.txt @ 1365:049b99f4b323
reply to OD
| author | Razvan Pascanu <r.pascanu@gmail.com> |
|---|---|
| date | Fri, 12 Nov 2010 11:49:00 -0500 |
| parents | 01157763c2d7 |
| children | f945ed016c68 |
comparison
equal
deleted
inserted
replaced
| 1364:01157763c2d7 | 1365:049b99f4b323 |
|---|---|
| 225 theano.function([index], graph(my_dataset)[index].variable) | 225 theano.function([index], graph(my_dataset)[index].variable) |
| 226 while with (1) the same function is compiled implicitly with: | 226 while with (1) the same function is compiled implicitly with: |
| 227 for sample in graph(my_dataset): | 227 for sample in graph(my_dataset): |
| 228 ... | 228 ... |
| 229 | 229 |
| 230 RP answers: right. I was actually constructing this stupid example in my mind when | |
| 231 you would do like : | |
| 232 i1 = f1(data) | |
| 233 i2 = f2(i1) | |
| 234 i3 = f3(i2) | |
| 235 ... | |
| 236 iN = fN(iN-1) | |
| 237 and then you would say .. wait I want to do this on new_data as well. Oh no, I | |
| 238 have to copy the entire block or whatever. That is so annoying. But actually you | |
| 239 could just write: | |
| 240 | |
| 241 def my_f(data): | |
| 242 i1 = f1(data) | |
| 243 ... | |
| 244 return iN | |
| 245 | |
| 246 and then just use that function which is what you pointed out. I agree I'm | |
| 247 not sure anymore on the point that I was trying to make. Is like if you are | |
| 248 a lazy programmer, and you write everything without functions, you can | |
| 249 argue that you like more (2) because you only pass the dataset at the end | |
| 250 and not at the beginning. But if (1) would have the replace function this | |
| 251 argument will fail. Though this only stands if you like don't want to make | |
| 252 a function out of your pipeline that takes the dataset as input, which now | |
| 253 that I think about it is pretty stupid not to do. Sorry for that. | |
| 254 | |
| 255 | |
| 230 - in approach (1) the initial dataset object (the one that loads the data) | 256 - in approach (1) the initial dataset object (the one that loads the data) |
| 231 decides if you will use shared variables and indices to deal with the | 257 decides if you will use shared variables and indices to deal with the |
| 232 dataset or if you will use ``theano.tensor.matrix`` and not the user( at | 258 dataset or if you will use ``theano.tensor.matrix`` and not the user( at |
| 233 least not without hacking the code). Of course whoever writes that class | 259 least not without hacking the code). Of course whoever writes that class |
| 234 can add a flag to it to switch between behaviours that make sense. | 260 can add a flag to it to switch between behaviours that make sense. |