Collective Re-Entry Classifiers

The article proposes a new method for teaching private classifiers, as well as a way to aggregate their forecasts as part of a committee. The training is based on the hypothesis of iterative re-entry of biological neural networks and uses the principal component method for its implementation. For private classifiers, the areas of competence are defined in the aggregate covering the training set of examples. It is shown that the iterative learning process converges in several steps, ensuring 100% recognition accuracy in the area of competence of the private classifier. The aggregation of forecasts is implemented according to the principle of maximum projection of the image onto its own subspaces of classes of private classifiers. Examples of the use of the committee of competent classifiers for the MNIST dataset are given. A continuous learning model of the classifier committee is proposed that is suitable for building self-learning recognition systems.