Commit graph

201 commits

Author SHA1 Message Date
fb22f64716 tweak semantics etc. 2017-07-21 19:45:58 +00:00
217a19110a fix case when no callbacks are given 2017-07-21 19:45:34 +00:00
4a108a10ae allow MomentumClip, SineCLR, WaveCLR in config 2017-07-21 19:43:57 +00:00
e7a6974829 yeah probably not 2017-07-12 09:07:22 +00:00
928850c2a8 lower process priority 2017-07-11 12:44:26 +00:00
9f8ac737db update mnist network 2017-07-11 12:11:47 +00:00
112e263056 fix code i forgot to test, plus some tweaks 2017-07-11 11:36:11 +00:00
7bd5518650 note to self on how to handle generators 2017-07-11 11:23:27 +00:00
e145f0422f more YellowFin bullshit
i give up on this optimizer. its design is flawed.
2017-07-11 11:22:36 +00:00
436f45fbb0 rewrite Ritual to reduce code duplication 2017-07-03 11:54:37 +00:00
85c9b3b5c1 finish(?) implementing YellowFin 2017-07-03 09:48:22 +00:00
c41700ab8d add momentum with gradient clipping
i may refactor out clipping later,
but it's too nice not have right now.
2017-07-03 09:44:16 +00:00
6a3f047ddc rename alpha to lr where applicable 2017-07-02 05:39:51 +00:00
9706aaabbb add WIP YellowFin optimizer implementation 2017-07-02 02:55:19 +00:00
d8bf6d1c5b easier logging of losses etc. 2017-07-02 02:53:31 +00:00
1b1184480a allow optimizers to adjust their own learning rate 2017-07-02 02:52:07 +00:00
22dc651cce move lament into core 2017-07-01 02:22:34 +00:00
7da93e93a8 move graph printing into Model class 2017-07-01 02:17:46 +00:00
1352de7006 merge the right commit this time 2017-07-01 01:15:16 +00:00
69786b40a1 begin work on multiple input/output nodes 2017-07-01 00:44:56 +00:00
a530558fc1 begin support for multiple input/output layers 2017-06-30 21:51:26 +00:00
a7c4bdaa2e remove dead line and punctuate comment 2017-06-30 21:13:37 +00:00
a4c1c5c04c update readme 2017-06-26 01:49:46 +00:00
c02fba01e2 various
use updated filenames.
don't use emnist by default.
tweak expando integer handling.
add some comments.
2017-06-26 00:16:51 +00:00
a770444199 shorten names 2017-06-25 22:08:07 +00:00
1a2a8702a0 add missing imports 2017-06-21 13:07:57 +00:00
c7193690c1 fix default ftml parameter 2017-06-21 12:31:41 +00:00
5a4b3d2934 add Follow The Moving Leader optimizer 2017-06-21 12:20:27 +00:00
9c0cbba3ff move epsilon out of square-root (big improvement) 2017-06-21 00:13:53 +00:00
9b85b49ee5 add minimal example to readme 2017-06-18 01:58:40 +00:00
264c3abd83 allow integer expando values 2017-06-18 01:58:34 +00:00
35f380c5d0 rename broken variables 2017-06-17 23:41:02 +00:00
a4081606f7 shuffle by default 2017-06-17 17:12:59 +00:00
cf1b7c1c13 various parameter tweaks 2017-06-17 16:46:39 +00:00
e5cea3f847 add SELU stuff 2017-06-17 16:45:50 +00:00
7e38f45a3f add cosine activation 2017-06-09 12:32:44 +00:00
d3c23912c1 add Softplus and LeCunTanh; use LeCunTanh 2017-04-23 18:58:33 +00:00
0332c2662b reword readme 2017-04-12 05:23:07 +00:00
5cb28eeef7 tweak and fix 2017-04-11 12:49:49 +00:00
c49e498aa0 add dropout and deterministic predictions 2017-04-11 10:39:42 +00:00
9b730b0516 reorganize some code 2017-04-11 10:11:08 +00:00
87ffa014ca update mnist example with new features 2017-04-11 04:48:53 +00:00
d08b5b91a1 convert target losses to log10 format 2017-04-11 04:47:30 +00:00
904423d631 add saturating activity regularizer 2017-04-11 04:46:54 +00:00
b6597e8b6c fix SGDR restart iteration and add WaveCLR 2017-04-11 04:46:40 +00:00
a448ff3e8a add weight regularization 2017-04-10 14:36:08 +00:00
0619163447 fix shape assertions 2017-04-10 10:34:58 +00:00
a78fc98215 greatly refactor weight handling 2017-04-10 10:08:41 +00:00
c3e2dd56bf refactor shape-handling a little 2017-04-10 08:26:38 +00:00
1d729b98aa fix epoch incrementing 2017-03-22 21:41:24 +00:00