One idea could be to find a way to use neural networks to know if a position can lead to a mate in a few moves or not. I don't mean that it is a "mate in N" but just "it could be a mate in a few moves". For sample, if you have a queen+bishop+knight near the enemy king, with few defenders, maybe you can mate easily. If the 1/8 rank is empty, just with a castled king and you have open columns and one or two rooks... it could becomes a mate.James Constance wrote:Hi all
I've just finished my second year of a Computer Science degree and need to come up with a topic for my final year's dissertation. I'd quite like to do something chess related, but I'm not sure what a suitable area to focus on would be. My supervisor has some expertise in pattern recognition, so I thought perhaps there might be something there.[...]
With an engine, you can easily find a mate in a few moves but you can't just look at the chessboard and say: "oh, the black will be mated soon", without a full calculation of all variants. This is why I think that a similar solution could be very hard to find but neural networks could help (and they must do a kind of pattern recognition, as requested).