Abstract

I consider the problem of learning concepts from small numbers of positive examples, a feat which humans perform routinely but which computers are rarely capable of. Bridging machine learning and cognitive science perspectives, I present both theoretical analysis and an empirical study with human subjects for the simple task of learning concepts corresponding to axis-aligned rectangles in a multidimensional feature space. Existing learning models, when applied to this task, cannot explain how subjects generalize from only a few examples of the concept. I propose a principled Bayesian model based on the assumption that the examples are a random sample from the concept to be learned. The model gives precise fits to human behavior on this simple task and provides qualitative insights into more complex, realistic cases of concept learning. 1

Citations

275	ALCOVE: An exemplar-based connectionist model ot category learning - Kruschke - 1999
232	Toward a universal law of generalization for psychological science - Shepard - 1987
180	Solving the multipleinstance problem with axis-parallel rectangles - Dietterich, Lathrop, et al. - 1997
179	A Study of Thinking - Bruner, Goodnow, et al. - 1956
98	Bounds on the Sample Complexity of Bayesian Learning Using Information Theory and the VC Dimension - Haussler, Kearns, et al. - 1991
56	A novelty detection approach to classification - Japkowicz, Myers, et al. - 1995
25	The structure of perceptual categories - Feldman - 1997
21	Fast learning by bounding likelihoods in sigmoid type belief networks - Jaakkola, Saul, et al. - 1996
1	Learning from positive data - Mitchell - 1997