| Age | Commit message (Expand) | Author |
|---|
| 39 hours | Better status | tslil clingman |
| 39 hours | Shed dependency of linenoise | tslil clingman |
| 39 hours | New weights from a wider training | tslil clingman |
| 39 hours | Beginning to work on probabilistic depth now that MCU is out | tslil clingman |
| 39 hours | Might as well use all of them... | tslil clingman |
| 39 hours | Went back up to 1986 weights | tslil |
| 39 hours | Quick sketch of non-determinism | tslil |
| 39 hours | Another road bug | tslil |
| 39 hours | Dragon is not separate, fixed concession, allowing playing as b/w | tslil |
| 39 hours | Changed function back to macro | tslil |
| 39 hours | There was a bug in the minimax! | tslil |
| 39 hours | Better trained weights, an important oversight in minimax (steps <=) | tslil |
| 39 hours | So after a small tweak, it actually plays well! | tslil |
| 39 hours | Seems tolerable, but not great. Still wont win or lose ??? | tslil |
| 39 hours | Trying alpha-beta pruning. There's a bug somewhere ... | tslil |
| 39 hours | There are still some bugs somewhere in minimax ... | tslil |
| 39 hours | Fixed minimax (!), fixed bugs in tak.c | tslil |
| 39 hours | Trying minimax | tslil |
| 39 hours | Crass error | tslil |
| 39 hours | Weights? | tslil |
| 39 hours | Wider? | tslil |
| 39 hours | Trying various things to teach ti | tslil |
| 39 hours | Have to account for CPS | tslil |
| 39 hours | Well it works, and it's bad. Many bugs fixed in interim | tslil |
| 39 hours | Closing in! | tslil |
| 39 hours | Auto-generating weights | tslil |
| 39 hours | Generate the correct format directly | tslil |
| 39 hours | New weights | tslil |
| 39 hours | one output (of course), Pade approximant of logistic for activation | tslil |
| 39 hours | Train on data that is close to the end of the game only | tslil |
| 39 hours | Working version of this script | tslil |
| 39 hours | This finally works! | tslil |
| 39 hours | Hopefully removed all layer bugs | tslil |
| 39 hours | Trying again with correct data generation | tslil |
| 39 hours | Also this was wrong | tslil |
| 39 hours | Close in principle, but there;s a bug in ct1997 &/ pptdb! | tslil |
| 39 hours | This is the general idea, but i'm sure order is incorrect somewhere | tslil |
| 39 hours | Starting on importing network! | tslil |
| 39 hours | Two inputs to model, stacks and flat counts | tslil |
| 39 hours | Fixed comments during extraction | tslil |
| 39 hours | Shuffling is important | tslil |
| 39 hours | Magic | tslil |
| 39 hours | Fixed a small bug in PTN parsing, other stuff | tslil |
| 39 hours | Closing in on something reasonable for roads | tslil |
| 39 hours | Check for road wins before flat wins | tslil |
| 39 hours | Searching for a good representation | tslil |
| 39 hours | Small tweaks | tslil |
| 39 hours | On the hunt for a better representation | tslil |
| 39 hours | Experimenting with training nn | tslil |
| 39 hours | Separate concerns in pptdb, remove legacy in extract.sh | tslil |
