Kx-trees: An unsupervised learning method for use in developmental agents

Acquiring concepts from experience is a key aspect of development and one that is commonly neglected in learning agents. In this work, concept acquisition is formulated as an unsupervised learning problem and is addressed with a novel algorithm: kx-trees. kx-trees differ from prior approaches to unsupervised learning in that they require very little information; four user selected parameters determine all aspects of kx-trees´ performance. Notably, and in contrast with most other unsupervised learning approaches, they do not require that the input state space be well-scaled. kx-trees´ operation is described in detail and illustrated with two simulations. The second simulation shows some similarities between kx-trees and feature construction in the human visual processing system.