ctak - An implementation of Tak and a computer opponent in C

Age	Commit message (Collapse)	Author
46 hours	new neural network arch (faster + better) & minor changes + fixes	tslil clingman
	Gone is the convolutional neural network, for it turns out not only is it more difficult to train, but all of the extra information about board layers didn't make much of a difference at this size. So cnn1986 has been replaced by nn1986, a standard, two-layer, dense nn configured as a binary classifier and (mis)used in that capacity. Note: total number of parameters is unchanged. HARK: this new nn exposes a bug somewhere in ctak. Run ctlm with self-play to see the completely borked board state at the end.
46 hours	Welcome geminict!	tslil clingman
	This is a special interface to negamax_cnn1986 which is designed to generate output for use in a CGI tak interface to be used over gemini. Also in this commit is a reformating of the various source files to use the traditional tab width of 8 spaces.
46 hours	Corrected generation of training data for 6s	tslil clingman

46 hours	Don't generate header for training data + tweaks	tslil
	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.
46 hours	Change the training data generation a little	tslil clingman
	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.
46 hours	TEI interface working!	tslil clingman

46 hours	Tried some naive iterative deepening. Work on TEI interface next	tslil clingman
	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.
46 hours	Just some #weightgoals ;)	tslil clingman
	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.
46 hours	Syntax errors, small tweak to training data generation	tslil clingman

46 hours	Added license information!	tslil clingman

46 hours	Added tunable search depth and self-play	tslil clingman

46 hours	Toying with symmetrising data	tslil clingman

46 hours	Went back up to 1986 weights	tslil

46 hours	Changed function back to macro	tslil

46 hours	Seems tolerable, but not great. Still wont win or lose ???	tslil

46 hours	Fixed minimax (!), fixed bugs in tak.c	tslil
	With minimax of depth 1 the evaluation function seems alright with the current method of training and generating weights
46 hours	Trying minimax	tslil

46 hours	Weights?	tslil

46 hours	Wider?	tslil

46 hours	Trying various things to teach ti	tslil

46 hours	Generate the correct format directly	tslil

46 hours	Train on data that is close to the end of the game only	tslil

46 hours	Hopefully removed all layer bugs	tslil

46 hours	Trying again with correct data generation	tslil

46 hours	Close in principle, but there;s a bug in ct1997 &/ pptdb!	tslil

46 hours	Two inputs to model, stacks and flat counts	tslil

46 hours	Shuffling is important	tslil

46 hours	Fixed a small bug in PTN parsing, other stuff	tslil

46 hours	Closing in on something reasonable for roads	tslil

46 hours	Searching for a good representation	tslil

46 hours	Small tweaks	tslil

46 hours	On the hunt for a better representation	tslil

46 hours	Experimenting with training nn	tslil

46 hours	Separate concerns in pptdb, remove legacy in extract.sh	tslil

46 hours	Gotta go fast	tslil

46 hours	First attempt at training data + exclude draws + list overflows too	tslil
	Take the output and do grep -Fvxf output_file input_file to drop the bad games from consideration
46 hours	Tabs for indentation, spaces for alignment	tslil

46 hours	Fixed indentation and some bugs, stats programme	tslil