基于概率的归纳推理模型  

Probability-based Models of Inductive Reasoning

作　　者：邢强[1] 车敬上 

机构地区：[1]广州大学心理学系,广州510006 [2]广东技术师范学院天河学院,广州510540

出　　处：《心理学探新》2012年第6期508-513,共6页Psychological Exploration

基　　金：国家自然科学基金(31070918);广州市属高校人文社会科学重点项目(10A108)

摘　　要：归纳推理是指从个别前提推出一般结论的推理,它是人的一项基本认知能力,也是心理学研究需要解释的重要现象。在模拟归纳推理时,不能排除先验知识对归纳推理结论力度的影响,贝叶斯理论可以结合条件对先验概率进行修正,能够很好地将先验知识引入到归纳推理中来。文章重点介绍了基于记忆的贝叶斯模型,相似性概率模型和结构统计模型对归纳推理的解释。Inductive reasoning can make powerful generalizations from sparse data. It is a basic capacity and a phenomenon that psychol- ogist should explain. The background knowledge plays an important role in inductive reasoning. So the computational models of inductive reasoning will be imperfect without knowledge. Bayesian theory can revise prior probability by combining the evidence. Bayesian infer- ence in probabilistic models can explain how prior knowledge is used in inductive reasoning. The paper introduces three probabilistic models of inductive reasoning： memory - based Bayesian model, SimProb model and structured statistical mode.

关 键 词：归纳推理 贝叶斯理论 概率模型 结构统计模型 

分 类 号：B842.5[哲学宗教—基础心理学]