Spring 2000 Machine Learning Othello Projects

• New Primitives
  • Leonid Portnoy - Investigated using primitives based on the X/Y location of the piece, and a primitive which guessed the value of the next board randomly.
• Fitness Measure
  • Thomas Keerikattu - Modified the fitness-measure algorithm so that fitness is evaluated against several players. The new algorithm uses the number of players defeated as a fitness measure and ignores the number of pieces remaining on the board. Used a reduced set of terminals: 'white' terminals are removed and 'constant' terminals modified.
  • George Yi, Naho Osagawara, and Wei-Ang Lee - Dare to Mutate: Small population after generations and generations of evolutions (adaptiving to random mutation events) give birth to one smart dude that can slightly overmatch Edgar. With this dude's structure left in tact, the "extremities" of the dude, tree edges, are refined to play Othello (as well as the internal nodes). Genetically engineered, the super dude creams Edgar, and many other random opponents called to compete against him. Resistance is futile.
• Time information
  • Peng Xu - a. Generally investigate how to use multiple trees to evaluate in GP. b. Two approaches to integrating time information into evaluation. 1)Multiply trees evaluation in different time points 2)New primitives that can determine the branch to evaluate according to current time points.