Pattern recognition using soft addressing techniques

Soft addressing techniques include those recognition methods which make it possible to use direct transformations from the pattern description to the memory address. The concept of soft addressing applied to pattern recognition problems is considered. Particular recognition methods are compared and discussed in relation to that concept. An approach to soft addressing which is based on the notion of a variable-dimensional feature space is suggested. It is shown that the approach can advance and improve recognition procedures which are similar to the nearest neighborhood method. Practical applications of the suggested approach to visual data processing are presented