BTW: maybe you could translate to java also normdiff function (to get the network derivatives), just to complete the tool... Obviously if you have time and if you interested in this...
Greetings,
Marek
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I left the normdiff function untranslated since I thought it only useful for training the network. If there is use for it, I will look into it. Can't say anything about the schedule, though.
br, Juhani
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hallo!
ffnet can calculate exact derivatives of the network outputs versus its inputs (one of the unique features of ffnet :)). This is done by normdiff. These values are definitely not the same derivatives which are used in training (where derivatives versus weights are used).
Greetings
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Adds Java export to FFnet
I forgot to add the new file _py2j.py to setup.py. The additional patch is here:
Index: setup.py
--- setup.py (.../tags/0.6.2/setup.py) (revision 8237)
+++ setup.py (.../trunk/setup.py) (revision 8237)
@@ -28,6 +28,7 @@
'ffnet._version',
'ffnet.fortran.__init__',
'ffnet.tools._py2f',
+ 'ffnet.tools._py2j',
'ffnet.tools.drawffnet'],
ext_modules = [ext1, ext2],
data_files = [('ffnet/examples', ['examples/xor.py',
Hi!
Thanks for patches. They are in trunk now.
BTW: maybe you could translate to java also normdiff function (to get the network derivatives), just to complete the tool... Obviously if you have time and if you interested in this...
Greetings,
Marek
Hi!
I left the normdiff function untranslated since I thought it only useful for training the network. If there is use for it, I will look into it. Can't say anything about the schedule, though.
br, Juhani
Hallo!
ffnet can calculate exact derivatives of the network outputs versus its inputs (one of the unique features of ffnet :)). This is done by normdiff. These values are definitely not the same derivatives which are used in training (where derivatives versus weights are used).
Greetings