Family of pattern classifiers

A good deal of argument has been devoted to the relative merits of various established classification schemes. In the letter, a number of these apparently different classifiers are shown to be special cases of a general classifier. The optimum performance obtainable from each is much the same with a given set of design data, so that the choice in any particular case depends largely on the cost of implementation. Significant improvements in performance might be obtainable by improving, or making better use of, the design data, but not by changing to a different classifier.