Spatial-Temporal Reasoning Applications of Computational Intelligence in the Game of Go and Computer Networks

Spatial-temporal reasoning is the ability to reason with spatial images or information about space over time. In this dissertation, computational intelligence techniques are applied to computer Go and computer network applications. Among four experiments, the first three are related to the game of Go, and the last one concerns the routing problem in computer networks. The first experiment represents the first training of a modified cellular simultaneous recurrent network (CSRN) trained with cellular particle swarm optimization (PSO). Another contribution is the development of a comprehensive theoretical study of a 2x2 Go research platform with a certified 5 dan Go expert. The proposed architecture successfully trains a 2x2 game tree. The contribution of the second experiment is the development of a computational intelligence algorithm called collective cooperative learning (CCL). CCL learns the group size of Go stones on a Go board with zero knowledge by communicating only with the immediate neighbors. An analysis determines the lower bound of a design parameter that guarantees a solution. The contribution of the third experiment is the proposal of a unified system architecture for a Go robot. A prototype Go robot is implemented for the first time in the literature. The last experiment tackles a disruption-tolerant routing problem for a network suffering from link disruption. This experiment represents the first time that the disruption-tolerant routing problem has been formulated with a Markov Decision Process. In addition, the packet delivery rate has been improved under a range of link disruption levels via a reinforcement learning approach.