A dependent feature is a feature that only makes sense when some other feature has a specific value. For example, in a database of motorvehicles 'number of doors' is not meaningful if the type of vehicle is 'motorcycle'. Many machine learning algorithms will perform more efficiently if they take into account this form of dependency, for example, in a genetic algorithm it is not helpful to mutate features that cannot have any effect.
Used in Chap. 9: page 126