We discuss the learning problem in a two-layer neural network. The problem is reduced to a system of linear inequalities, and the solvability of the system is discussed.
