RELATIONS BETWEEN EXEMPLAR-SIMILARITY AND LIKELIHOOD MODELS OF CLASSIFICATION
RELATIONS BETWEEN EXEMPLAR-SIMILARITY AND LIKELIHOOD MODELS OF CLASSIFICATION
复制标题
DOI：
10.1016/0022-2496(90)90020-a
复制
发表时间：
1990-12-01
影响因子：
1.8
通讯作者：
NOSOFSKY, RM
中科院分区：
文献类型：
--
作者：
NOSOFSKY, RM
识别的相似性选择模型（Luce，1963；Shepard，1957）和分类的上下文模型（Medin 和 Schaffer，1978；Nosofsky，1986）被证明与各种基于可能性的模型密切相关。特别是，它表明：对于在独立维度上定义的类别分布，上下文模型的一般版本和 Estes''（1986）相似性似然模型在形式上是相同的；上下文模型和相似性选择模型可以被解释为基于样本的似然模型；独立特征增删模型是相似选择模型的特例；基于感知/可能性的识别决策模型生成可通过相似性选择模型表征的预测。
The similarity choice model of identification (Luce, 1963; Shepard, 1957) and the context model of categorization (Medin and Schaffer, 1978; Nosofsky, 1986) are shown to be closely related to a variety of likelihood-based models. In particular, it is shown that: for category distributions defined over independent dimensions, general versions of the context model and Estes'' (1986) similarity-likelihood model are formally identical; the context model and similarity choice model can be given interpretations as exemplar-based likelihood models; an independent feature addition-deletion model is a special case of the similarity choice model; and a perception/likelihood-based decision model of identification generates predictions that are characterizable by the similarity choice model.