A Human Like Incremental Decision Tree Algorithm: Combining Rule Learning, Pattern Induction, and Storing Examples

Early machine learning research was strongly interrelated with research on human category learning while later on the focus shifted to the development of algorithms with high performance. Only recently, there is a renewed interest in cognitive aspects of learning. Machine learning approaches might be able to model and explain human category learning while cognitive models might inspire new, more human like, approaches to machine learning. In cognitive science research there exist different theories of category learning, especially, rule-based approaches, prototypes, and exemplar-based theories. To take account of the flexibility of human learning and categorization we propose a human like learning algorithm. In the algorithm we combine incremental decision tree learning, least general generalization, and storing examples for similarity-based categorization. In this paper we present first ideas of this algorithm.