Complexity of Connectionist and Constraint-Satisfaction Networks

Abstract : Since the beginning of the funding of the grant, we established a substantial effort in the area of connectionist optimization algorithms, relaxation networks, and geometrical learning algorithms. All of the above are highly interconnected research projects. We have achieved several significant results that have increased our understanding of the computational capabilities and limitations, of connectionist and constraint network. Our most significant contributions thus far are in the area of parallel complexity of constraint networks. comparative experimentation with learning algorithms and geometric concept learning. Our results in the area of parallel constraint networks are the subject of several publication in first rate journal and conferences. Our experimental research achieves the best results on several well established benchmarks. Most notably our group achieved the best results (in terms of predictions accuracy) in the area of protein folding. The technical result of research investigations are summarized in the following sections.