Abstract: For a class of multi-input-multi-output (MIMO) discrete-time nonlinear systems with triangular form control inputs, an adaptive neural network control is proposed via backstepping. Due to the non-affine form of the control inputs, feedback linearization method can not be used to design control system. Therefore, implicit function theorem is firstly exploited to assert the existence of the ideal control inputs, which can compel the system outputs to track their desired trajectories, and then ideal control inputs are constructed. By using high-order neural networks to approximate the ideal control inputs, an adaptive neural network control is developed via backstepping design. All signals of the closed-loop system are proved to be semi-globally uniformly ultimately bounded under the proposed control, while the effectiveness of the proposed control is illustrated by simulations.