Logical-Combinatorial-Probabilistic Recognition Algorithms

摘要

Effective logical-combinatorial-probabilistic algorithms of pattern recognition in the intelligent system with matrix knowledge representation are proposed. Algorithms of three types are constructed on the basis of all possible minimal (partly irredundant), unconditional, or mixed diagnostic testes. Final pattern recognition is performed according to the principle of voting on the set of all tests and algorithms. Algorithms are recommended for a small number of realizations of each pattern and a large set of features. The algorithms were partially implemented and tested in the system IMSLOG in the solution of actual problems. Positive results were achieved.