Evolutionary Neural Networks [中文]

Introduction: Many different techniques have been used in Artificial Intelligence (AI) research. Two of the most popular are evolutionary programming and neural networks. Both of these techniques have been well developed in the forty years since they were first proposed. In 1999, Dr. Fogel of the University of Southern California combined these two techniques in a way that he called: "Evolutionary Neural Networks". He demonstrated the power of this technique in developing a program that learned to play checkers without any expert knowledge.

Long term research objective: To apply this technique in the areas of bio-infomatics, decision making in telemedicine and natural language processing.

Short term research objective: To study the performance of this algorithm and demonstrate its ability to solve standard AI problems in the field of game theory.

Research Plan: (Click on the following flash diagram)

Past: technique applied to training perceptons, and then to Othello (a simple game that we can use absolute standards to evaluate play)
Current: Application to Chinese Chess (a more complicated game)
Next steps (available Master's Projects):
- Generalization of the code used in playing Othello and investigation of the average lifetime of a network, etc...
- Combine evolution of the lookahead function with evolution of the evaluation function. Game: Othello.
- Finish application to Chinese Chess
- Application to Japanese Go world problems requiring pattern recognition.

Evolutionary Neural Networks [中 文]

Evolutionary Neural Networks [中文]