another question on romi learning

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

ozziejoe
Posts: 811
Joined: Wed Mar 08, 2006 10:07 pm

another question on romi learning

Post by ozziejoe »

this is probably a real basic question, but...

Does romi learning recognize transpositions? e.g., Path A gets you to a particular position, and romi learns. Then another path B gets you to the same position. can romi use what she learned?

Romi continues to hit above her weight, thanks to learning (see below). Romi should be losing to every one of htese engines.

I put rybka in there no so romi can beat her but so romi can learn from her

1: Romichess 74.5/165 ····························
2: Rybkav2.3.2a.w32 25.0/27 1=1111==111111111111=111111
3: Slow 17.0/27 11111===0111=01010=011101=0
4: Ruffian_210 15.5/28 11=10=0100=101=1=1=10101==0=
5: List512 13.5/28 =011110001010100010101010101
6: Yace Paderborn 10.5/28 01011=0==10=0=010100==0=000=
7: SOS 5.1 for Arena 9.0/27 ==000000111=0011=1000000100
Michael Sherwin
Posts: 3196
Joined: Fri May 26, 2006 3:00 am
Location: WY, USA
Full name: Michael Sherwin

Re: another question on romi learning

Post by Michael Sherwin »

I tried to implement tranpositions using the current file structure, but failed.

The now mythical rewrite that I have been working on forever will recognise transpositions if/when it ever enters the realm of reality.

Edit: Romi usually enters a slump along the way as she is 'fishing' for effective new lines. Then when she finds them her performance shoud take a jump in the positive direction. It can take awhile though.
If you are on a sidewalk and the covid goes beep beep
Just step aside or you might have a bit of heat
Covid covid runs through the town all day
Can the people ever change their ways
Sherwin the covid's after you
Sherwin if it catches you you're through
ozziejoe
Posts: 811
Joined: Wed Mar 08, 2006 10:07 pm

Re: another question on romi learning

Post by ozziejoe »

Hi michael,

If you could solve the transposition issue, I think that this would add a great deal of strength to the learning.

I am currently designing a "narrow" romi book that focuses on positions where white plays 1 nf3.....2. c4.....3 g3.... The idea is to vastly reduce the lines that romi plays in order to have her learning come into play. Nf3 is highly transpositional, so if romi could recognize transpositions, it would really be able to use its knowledge...

Oh, and by the way. Romi just figured out how to beat Rybka in the one opening. It got its first win after 28 games (as black). However, it is playing black again now, and losing. It looks like it changed its mind on what to play, alas...It does have to come up with lines that work against everybody.

best
J