Symbolic learning algorithms, although conceptually simple, have been shown to be effective in some applications that are traditionally dominated by other learning methods, such as speaker recognition ([Squ94]). They have several advantages that would appear to make them a good candidate for applications in gesture recognition -- such as advantages in speed of evaluation and computing power required, in terms of time and space.
Similarly, investigations of the use of instance-based learning techniques have been limited in scope, but are one of the most trivial forms of learners available.