Pattern recognition in chess using unsupervised learning
11 / 1999 - 11 / 2003
The research proposed aims at combining computer and human expertise in building a strong artificial chess player. Previous attempts stessed the incorporation of specialised human knowledge into chess programs, but they failed. Computational models of chess programming have proven to be extremely powerful. Now we feel the time to be ripe to re-address the focus on human chess models and attempt to develop a combined human-computer model for chess programs in non-tactical domains. The research concentrates on extracting chess patterns from test positions and game scores using specialised neural networks architectures for unsupervised learning.