Search found 3452 matches

by Henk
Mon Sep 17, 2018 11:16 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

Joost Buijs wrote:
Mon Sep 17, 2018 11:12 am
Why don't you switch to ReLU as activation function, at least it is fast because it doesn't need exp().
Relu is not self normalizing. If I use RELU I have to implement batch normalization as wel. Might be that adding batch normalization might be making it more slow. I don't know yet.
by Henk
Mon Sep 17, 2018 10:47 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

For loss function I still use mean square error. Maybe I should change that. For activation function I use SELU. I read that if you use a SELU you don't need batch normalization for it is self normalizing. But computing an Exp(x) is one of slowest operations during training. static public double SE...
by Henk
Fri Sep 14, 2018 5:38 pm
Forum: Computer Chess Club: General Topics
Topic: Slipper: update (not much progress)
Replies: 1
Views: 93

Slipper: update (not much progress)

Promoting into a rook, 1. c2-c4 { 2.003 0.02 } { } 1. d7-d6 { 2.008 -0.04 } { } 2. Ng1-f3 { 1.973 0.02 } { } 2. f7-f6 { 1.982 -0.08 } { } 3. Qd1-c2 { 1.958 0.02 } { } 3. Ng8-h6 { 1.948 -0.07 } { } 4. Nb1-c3 { 1.903 0.04 } { } 4. Nh6-g4 { 1.902 -0.09 } { } 5. Nc3-e4 { 1.874 0.07 } { } 5. Ng4-h2 { 1.8...
by Henk
Wed Sep 12, 2018 6:31 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

After three days tuning. Using 100000 examples. Still playing bad. 0.152658045592678 0.146846164052304 0.152672921563584 0.146796660198495 0.152628787384212 0.146750211537819 0.152533272772998 0.146706348167116 1. Ng1-f3 { 2.005 0.13 } { } 1. f7-f6 { 2.008 -0.19 } { } 2. c2-c4 { 1.975 0.12 } { } 2. ...
by Henk
Mon Sep 10, 2018 2:21 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

Getting from 0.1738 to 0.1722 costing 49 minutes. [Last week I lost the weights because I accidentally generated a new network overwriting the old one. So I had to start all over again. Busy two days now getting to this] 0.173843260482675 0.155715351217672 0.17376476122923 0.155656586818636 0.173660...
by Henk
Tue Sep 04, 2018 10:41 pm
Forum: Chess Players Forum
Topic: Fun
Replies: 0
Views: 177

Fun

by Henk
Sat Sep 01, 2018 11:14 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

Hi hi hi. It might even be true that a network with bad learning properties or with bugs performs best for it generalizes better.
by Henk
Sat Sep 01, 2018 11:02 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

There are many network architectures. Why should a residual network be the best. Perhaps great many variants to try. But unfortunately testing them takes so much time. So my strategy is to start with smallest ones first. For it should take less time to test or to find out it is not working and also ...
by Henk
Sat Sep 01, 2018 10:02 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Learning time growing exponentially with number of training examples
Replies: 34
Views: 1214

Re: Learning time growing exponentially with number of training examples

Slow progress. 70000 examples and each epoch taking ten minutes. When progress stops I change network and it starts all over again. Maybe adding more training examples might help if progress stops. I don't know. Haven't tested that. Or perhaps I better use 10000 examples or less to evaluate a networ...
by Henk
Sat Sep 01, 2018 9:38 am
Forum: Computer Chess Club: Tournaments and Matches
Topic: CCCC is running
Replies: 12
Views: 769

Re: CCCC is running

What is missing is live comments and youtube livestream.