The Base Selection Task in Analogical Planning

Analogical planning provides a means of solving problems where other machine learning meth­ ods fail, because it does not require numer­ ous previous examples or a rich domain the­ ory. Given a problem in an unfamiliar domain (the target case), an analogical planning sys­ tem locates a successful plan in a similar do­ main (the bast case), and uses the similarities to generate the target plan. Unfortunately, the analogical planning process is expensive and in­ flexible Many of the limiting factors reside in the base selection step, which drives the anal­ ogy formation process. This paper describes two ways of increasing the effectiveness and ef­ ficiency of analogical planning. First, a par­ allel graph-match base selection algorithm is presented. A parallel implementation on the Connection Machine is described and shown to substantially decrease the complexity of base selection. Second, a base-case merge algorithm is shown to increase the flexibility of analogi­ cal planning by combining the benefits of sev­ eral base cases when no single plan contributes enough information to the analogy. The effec­ tiveness of this approach is demonstrated with examples from the domain of automatic pro­ gramming. 1 Introduction Analogy is a powerful planning tool. Engineers and sci­ entists rarely attack a problem in an unfamiliar domain from scratch. Instead, then rely on their experience with solving problems in similar domains. They adapt known techniques, map constraints from a solved problem to the new problem, and modify existing solutions to fit the current problem specification. Given a novel problem (the target case), an analogical planner selects a similar, solved problem (the base case), computes a mapping be­ tween the base and target problem descriptions, and uses the mapping the adapt the base solution to the current domain. When examples are lacking and domain theory is scarce, the intelligent agent draws on past experience in similar situations to attack a new problem. Although analogy is considered to be a powerful tool for machine planning, it is also viewed as an expensive task which is rarely applied to large-scale problems. As Buchanan states, analogical reasoning is a "pipe dream when matched against the harsh standards of robust­ ness of commercial applications" [Buchanan, 1990]. A second limitation of analogical planning systems is flexi­ bility. In inductive learning systems, when the learner is not performing well enough more examples can be input to improve the hypothesis. In analogical learning sys­ tems, the …