Abstract:
This article considers the problem of restoring the threshold function in the information protection node from a input and output in the case when not all values are known. To solve this problem, it is proposed to use a geometric algorithm for characterising a partially known threshold $k$-valued function. The article proves the convergence of the algorithm at the final step; it is also shown that as a result of the algorithm, a certain threshold function will be constructed, which will coincide with this function at all known points.
Keywords:Algorithm of learning of threshold functions; proof of convergence; threshold function; expansion coefficients; increase coefficients.