Pattern Classification

Pattern classification is a general information processing task.  In this task, a system is presented an example pattern, and must name the kind of pattern that it is (that is, it must assign the example pattern to a general category).  For instance, a perceptron might be trained to be a pattern classifier that can be presented X’s in different fonts, and classify each of these as being an X, and at the same time be presented other letters (in different fonts too), and classify each of these as not being an X.  Pattern classification is important to cognitive science because one can measure the power of an information processor in terms of how complex its ability to perform this task is.  For example, multilayer perceptrons can solve any pattern classification problem (they are arbitrary pattern classifiers (Lippmann, 1989)) and therefore can be argued to have the same power as a Universal Turing Machine .