Speeding up problem solving by abstraction: a graph oriented approach

摘要: Abstract This paper presents a new perspective on the traditional AI task of problem solving and techniques abstraction refinement. The is based well-known, but little exploited, relation between finding path in graph two given nodes. oriented view suggests families techniques, algebraic STAR abstraction. first shown to be extremely sensitive exact manner which problems are represented. abstraction, by contrast, very widely applicable leads significant speedup all our experiments. reformulation refinement as algorithms several enhancements, including an optimal algorithm, one radically technique: alternating search direction . Experiments comparing these variety show that opportunism (AltO) variant technique, uniformly superior others.