Analogical and inductive reasoning

摘要: The first problem discussed in this thesis is the logical of analogy, which, given a formal definition analogical inference, asks under what conditions such inferences may be justified. By showing inadequacy approaches based on degree similarity between analogues, importance relevance known and inferred similarities highlighted. need for semantics motivates determinations, first-order expressions capturing idea predicate schemata. Determinations are shown to justify single-instance generalizations non-trivially, express an apparently common form knowledge hitherto overlooked knowledge-based systems. Analogical reasoning implemented MRS, logic programming system, more efficient than simple rule-based methods some important inference tasks. ability acquire use determinations strictly increase system can make from set data. Programs described inductive acquisition their subsequent aid construction large base. The second problem, suggested by subsuming first, identify ways which existing used help learn experience. A method enumerating types knowledge, but one, that contribute learning, general machine these ideas. The application logical, approach problems analogy induction thus shows able detect as many forms regularity possible order maximize its inferential capability. possibility aspects 'common sense' captured complex, abstract regularities suggests further empirical research knowledge.