ctak/src, branch sixes An implementation of Tak and a computer opponent in C https://git.l-3.space/ctak/atom/src?h=sixes 2021-03-28T03:59:33Z Playing around with huge nn's, poorly trained. 2021-03-28T03:59:33Z tslil tslil@posteo.de 2021-03-28T03:59:33Z urn:sha1:4cb4d7a24c3bcd263ce6fb4b83ebbde017650ed2 Not great at shallow depths Don't generate header for training data + tweaks 2021-02-12T02:04:42Z tslil tslil@posteo.de 2021-02-12T02:04:42Z urn:sha1:dd25d66353121939a8f9b2bfe6a20f4100ff4976 For some reason it would seem that moving flats to a lower value and increasing the proximity between caps and top flats improves acquisition. Still not great, but every bit counts. Change the training data generation a little 2021-02-11T01:07:24Z tslil clingman tslil@posteo.de 2021-02-11T01:07:24Z urn:sha1:52b2268a1c7b97809bdf0f02147f885ddfa6d397 Although it pains me to say it, ``label smoothing'' appears to be actually work. I'm also currently experimenting with training simply against _all_ games, instead of only bot matches. Once the training finishes i'll pit cttei against itself with old and new weights, hopefully there'll be a noticeable improvement. Rename ct_k -> ct, IANAL but ... 2021-02-05T21:49:14Z tslil tslil@posteo.de 2021-02-05T21:49:14Z urn:sha1:5c8e082276a24b9f266396874d3e72c05990a3df Small typo in generated output for weights 2021-02-05T21:37:32Z tslil clingman tslil@posteo.de 2021-02-05T21:37:32Z urn:sha1:105edbb2ccb6d902abb8bc3f8a5add896e5f161e Correct line clearing behaviour, EXIT_FAILURE <~ -1 2021-02-05T21:27:40Z tslil clingman tslil@posteo.de 2021-02-05T21:24:39Z urn:sha1:2122a979fefd27112720a47ecf585933fd9bcde4 Renamed binaries 2021-02-05T19:02:36Z tslil clingman tslil@posteo.de 2021-02-05T19:02:36Z urn:sha1:ed070ebc252f6d6d392b51b3a836bd08965b3946 TEI interface working! 2021-02-05T05:12:18Z tslil clingman tslil@posteo.de 2021-02-03T00:23:54Z urn:sha1:4f20d9fd63b6b3e30130c84c20f7cdbbbf2df76d Tried some naive iterative deepening. Work on TEI interface next 2021-02-02T02:46:24Z tslil clingman tslil@posteo.de 2021-02-02T02:46:24Z urn:sha1:41b5dc810e25724e8111cf7f2ad1b4b52f9b1206 If TEI is implemented, then i could make use of Morten's racetrack (https://github.com/MortenLohne/racetrack) and develop a quantitative measure of the bot's performance. This is the current priority. Just some #weightgoals ;) 2021-02-01T04:33:39Z tslil clingman tslil@posteo.de 2021-02-01T04:33:39Z urn:sha1:a7d80d7346a7c6944d917df4903342527b17d71c It turns out that while i was training on a 0/1 classification problem, i was using 2*eval - 1. Training using this function instead, and on bot-dominated game choices (chosen_player in extract.sh) seems to have given a better evaluation function. At the least, Morten's swindle doesn't work anymore.