This note presents an H-infinity design approach for a neural net-based control scheme. In this scheme, a class of nonlinear systems is approximated by two multilayer perceptrons. The neural networks are piecewisely interpolated to generate a linear differential inclusion model. Based on this model, a state feedback control law is designed. The H-infinity control is specified to eliminate the effect of approximation errors and external disturbances to achieve desired performance. It is shown that finding the permissible control gain matrices can be transformed to a standard linear matrix inequality (LMI) problem and solved using the convex optimization method.