Concept learning is when the goal of learing is to allocate input into one of a number of distinct classifications. Often concept learning is applied to a binary condition: 'in category' vs 'not in category', and some algorithms are typically expressed in this binary form. Examples of concept learning include symbolic algirthms such as version spaces and ID3, and also sub-symbolic technqies including many kinds of neural network and swarm computing.
Used in Chap. 3: page 31; Chap. 5: pages 63, 65; Chap. 16: pages 240, 248; Chap. 18: page 282