Chapter 8 - Learning

Abstract

This chapter provides an overview of the learning process in humans and machines. Learning can be thought of as adaptation to the environment based on experience. This inevitably requires new knowledge, new skills, or the reorganization of existing knowledge. The act of learning is motivated by attempts to improve a system by enabling better performance or avoiding poor performance. Machine learning involves: acquiring new information and knowledge, acquiring new skills, and finding new ways of organizing existing knowledge. When it is built, a machine will have a certain amount of information and knowledge designed into it. To learn it must also have some meta-knowledge built in. In particular, the machine must be able to absorb new data and operate on them so that they can be used in a purposeful way. This assumes that the machine is able to store this accumulating knowledge in appropriate data structures, that it has techniques for transforming raw data from its sensors into knowledge, and that it is able to manage its information base. This chapter explains basic concepts related to learning by memory, learning by updating parameters, Bayesian learning, learning from examples, learning by analogy, and learning by observation and discovery.