Mathematically, pattern recognition is a classification problem. Consider the recognition of characters. We wish to design a system such that a handwritten symbol will be recognized as an “A,” a “B,” etc. In other words, the machine we design must classify the observed handwritten character into one of 26 classes. The handwritten characters are often ambiguous, and there will be misclassified characters. The major goal in designing a pattern recognition machine is to have a low probability of misclassification.