Titlebook: Changes of Problem Representation; Theory and Experimen Eugene Fink Book 2002 Springer-Verlag Berlin Heidelberg 2002 Extension.Performance.

Amylase

Extended Strips DomainThe . world is larger than the other domains; it includes ten object types and twenty-three operators (Figures 3.38-3.41). We have tested .. with two linear and one nonlinear gain functions (Figure 12.1).

失望昨天

Logistics DomainThe last series of experiments is based on the Logistics Domain [Veloso, 1994], which includes eight object types, six operators, and two inference rules (Figure 3.43). We have experimented with three linear gain functions and three nonlinear functions (Figure 13.1).

截?cái)?/a>

Concluding remarksThe main contribution of the reported work is an architecture for improving representations in AI problem solving. The architecture is based on two principles proposed by Simon during his studies of human problem solving:

GLOOM

Motivationwe describe it. Researchers in psychology, cognitive science, and artificial intelligence have accumulated much evidence on the importance of appropriate representations for human problem solvers and AI systems.

Lipoma

Prodigy searchir technique combined goal-directed reasoning with forward chaining from the initial state. The authors of later systems [Fikes and Nilsson, 1971; Warren, 1974; Tate, 1977] gradually abandoned forward search and began to rely exclusively on backward chaining.

Conspiracy

Multiple representationsSection 6.1). Then, we propose a general model of using multiple representations (Section 6.2) and evaluating their utility (Section 6.3). Finally, we discuss the assumptions that underlie the developed model (Section 6.4).

貧窮地活

Statistical selectionon (Section 7.1), derive a solution to this problem (Sections 7.2 and 7.3), use it in choosing a representation (Sections 7.4), and give empirical results that confirm the effectiveness of the selection algorithm (Sections 7.5 and 7.6).

夜晚

Statistical extensionsen problem, we can utilize it to make a more accurate selection. We describe the use of problem-specific gain functions (Section 8.1), estimated problem complexity (Section 8.2), and similarity among problems (Section 8.3).

暴露他抗議

揭穿真相