Rational Pattern Classification When Very Little is Known

The problem of pattern classification when very little is known about the pattern source is structured through a set of axioms describing desirable properties for a classifier that involve no probabilistic assumptions or hypotheses. The derived classifier is then studied under the hypothesis that the pattern source is probabilistic and its long-run performance found to strongly converge to that of the Bayes classifier under reasonable restrictions. The sparse results of a pilot study on the comparative small sample performance of the derived classifier are found to be encouraging but not definitive. This work is an outgrowth of previous work on extrapolation when little is known