Supervised Classification and Finding Frequent Elements in Data

Correct logical classifiers the construction of which uses the ideas underlying the well-known “Kora” algorithm and the ideas of estimates calculating algorithms are considered. A modification of the training stage that in certain cases makes it possible to reduce the time cost without loss in the classification quality is proposed. Results of experiments on model and real-life data are presented, and new theoretical estimates of the complexity of training Kora-type classifiers are obtained.