This dissertation describes, demonstrates, and analyzes a new search technique called stochastic iterated genetic hillclimbing (SIGH). SIGH performs function optimizations in high-dimensional, binary vector spaces. Although the technique can be characterized in abstract computational terms, it emerged from research into massively parallel connectionist learning machines, and it has a compact implementation in a connectionist architecture. The behavior generated by the machine displays elements of two existing search techniques—stochastic hillclimbing and genetic search.
KeywordsSearch Strategy Global Maximum Strong Constraint Graph Partitioning Weak Constraint
Unable to display preview. Download preview PDF.