A criterion based on an information theoretic measure for goodness of fit between classifier and data baseA criterion for characterizing an iteratively trained classifier is presented. The criterion is based on an information theoretic measure that is developed from modeling classifier training iterations as a set of cascaded channels. The criterion is formulated as a figure of merit and as a performance index to check the appropriateness of application of the characterized classifier to an unknown data base and for implementing classifier updates and data selection respectively.
Document ID
19740037368
Acquisition Source
Legacy CDMS
Document Type
Conference Proceedings
Authors
Eigen, D. J. (Bell Telephone Laboratories, Inc. Murray Hill, N.J., United States)
Davida, G. I.
Northouse, R. A. (Wisconsin, University Milwaukee, Wis., United States)