Search found 3885 matches

by syzygy
Tue Aug 21, 2018 2:14 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: 7-men Syzygy attempt
Replies: 505
Views: 57548

Re: 7-men Syzygy attempt

These things tend to be exponential. In particular, you'd need a machine with a lot more RAM than 1TB. I thought roughly speaking you need 60 times more hard disk and ram for each extra piece. I’m sure the factor should be much smaller than 60. Long time ago someone has already predicted that build...
by syzygy
Sat Aug 18, 2018 10:30 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Re-Pair compression questions
Replies: 3
Views: 345

Re: Re-Pair compression questions

3) How much more reduction do you get from Huffman? I don't know this either. OK, I thought perhaps you know the (weighted) average symbol length of your Huffman coding since you store the Huffman encoding in the db, right? For each symbol length I store the number of symbols with that length. This...
by syzygy
Sat Aug 18, 2018 11:59 am
Forum: Computer Chess Club: General Topics
Topic: The poor evaluation function of lc0
Replies: 3
Views: 395

Re: The poor evaluation function of lc0

It might not be able to draw this with white, so -12.44 might be accurate from that point of view.
by syzygy
Fri Aug 17, 2018 6:49 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Re-Pair compression questions
Replies: 3
Views: 345

Re: Re-Pair compression questions

A few questions to Ronald about the Re-Pair compression algorithm (as preparation for draughts database compression). From my understanding of other posts here, your compression does a) a re-pair until either no pair has count > 1, or 4096 symbols have been generated, b) you follow this with a Huff...
by syzygy
Wed Aug 15, 2018 10:12 pm
Forum: Computer Chess Club: General Topics
Topic: Copyright and Machine Learning IP
Replies: 102
Views: 1982

Re: Copyright and Machine Learning IP

You've described a process where a network has learnt "game rules" by observation of those rules in action. But I don't get the bit where the "rule creation creativity" was passed across to the NN. I would get it if you said the "rules" were passed across. Because, now the NN encodes the rules, but...
by syzygy
Wed Aug 15, 2018 7:57 pm
Forum: Computer Chess Club: General Topics
Topic: Copyright and Machine Learning IP
Replies: 102
Views: 1982

Re: Copyright and Machine Learning IP

syzygy wrote: So my point is that the set of weights will be just as (un)copyrighted as the equivalent source or object code. The extra mechanical transformation (the "direct map") makes no difference. Yes, I agreed with you earlier and switched to: Chrisw wrote: Agreed. For this particular ingenio...
by syzygy
Wed Aug 15, 2018 7:10 pm
Forum: Computer Chess Club: General Topics
Topic: Copyright and Machine Learning IP
Replies: 102
Views: 1982

Re: Copyright and Machine Learning IP

And you argue, I think, that this transformation is functional. No, I'm saying that the transformation is mechanical. Converting a book from ASCII to Morse code is a mechanical and can therefore not result in the addition of creative elements. Mechanically converting an NN weight into a few lines o...
by syzygy
Wed Aug 15, 2018 6:43 pm
Forum: Computer Chess Club: General Topics
Topic: Copyright and Machine Learning IP
Replies: 102
Views: 1982

Re: Copyright and Machine Learning IP

And you argue, I think, that this transformation is functional. No, I'm saying that the transformation is mechanical. Converting a book from ASCII to Morse code is a mechanical and can therefore not result in the addition of creative elements. Mechanically converting an NN weight into a few lines o...
by syzygy
Tue Aug 14, 2018 10:30 pm
Forum: Computer Chess Club: General Topics
Topic: Copyright and Machine Learning IP
Replies: 102
Views: 1982

Re: Copyright and Machine Learning IP

Well, the architecture of a neural net maps directly to a computer program. You could say that the net inputs (chess position eg) are variables, the fixed weights as program constants and the net architecture as the program. If you really wanted, you could unroll the program loops and insert the co...
by syzygy
Tue Aug 14, 2018 8:50 pm
Forum: Computer Chess Club: General Topics
Topic: Copyright and Machine Learning IP
Replies: 102
Views: 1982

Re: Copyright and Machine Learning IP

Well, the architecture of a neural net maps directly to a computer program. You could say that the net inputs (chess position eg) are variables, the fixed weights as program constants and the net architecture as the program. If you really wanted, you could unroll the program loops and insert the co...