We give a direct proof of the fact that for broadcast channels the maximal-error capacity region is equal to the average-error capacity region. Instead of making an excursion to stochastic encoders as in Csiszar and Korner [1981] we base pour proof on matrix-partitioning.
PUBLICATION:
In Problems of Control and Information Theory, Vol. 19 (4), pp. 339 - 347, (1990). A scanned version of this paper is available here.