SAMPLES FORMATION FOR PROBLEM SOLVING OF FEATURE CLASSIFICATION
The problem of sample formation to automate object classification on features is solved. The new method of training sample formation is offered. It maintains the retention of topological properties of an original sample in generated sub-sample and doesn’t require downloading an entire sample to a computer memory. It is reduces the sample size and lowers the resource requirements to a computer.
Keywords: sampling, selection of occurence, data reduction, classification, pattern recognition, data dimension reduction.