Two heuristics for arranging the order of feature-extraction operations in recursive bayesian decision rule

摘要

The recursive Bayesian decision rule has been widely used in pattern classification. This paper presents two heuristics, power of discrimination and power of justification, for arranging the order of feature-extraction operations in the Recursive Bayesian decision rule. The heuristics predict the performance of each operation and select the best one as the next operation. The experimental result shows that these methods are very promising.