An iterative method for computing the channel capacity of both discrete and continuous input, continuous output channels is proposed. The efficiency of new method is demonstrated in comparison with the classical Blahut - Arimoto algorithm for several known channels. Moreover, we also present a hybrid method combining advantages of both the Blahut - Arimoto algorithm and our iterative approach. The new method is especially efficient for the channels with a priory unknown discrete input alphabet.
Bibliographical noteFunding: Grant of Ministry of Education and Science of the Russian Federation
(14.B25.31.0003); EPSRC project UNLOC, the Government program of the leading research schools NSh-4447.2014.2