Construction of fault-tolerant signal feature subsets

New kinds of regularities in knowledge, viz., fault-tolerant signal-feature subsets, are proposed for use in intelligent test-pattern-recognition systems; the change in the values of these features indicates the transition of objects from one pattern to another. The algorithms for finding fault-tolerant signal-feature subsets are described and examples of their work are given. The algorithms for finding new regularities are implemented in a subsystem incorporated into the IMSLOG intelligent software tool based on test methods of pattern recognition. Examples from various problem domains are given to find subsets of signal features that are tolerant to measurement errors. The usability of these subsets is substantiated.