A bead is intended to reach a specified target point in the minimum-time when it travels along a certain curve on a vertical plane with the gravity. This is called the brachistochrone problem. Its minimum-time control input may be found using the calculus of variation. However, the accuracy of its minimum-time control input is not high since the solution of the control input is based on a table form of inverse relations for some complicated nonlinear equations. To enhance the accuracy, this paper employs the neural network to represent the inverse relation of the complicated nonlinear equations. The accurate minimum-time control is possible with the interpolation property of the neural network. For various final target points, we have found that the proposed method is superior to the conventional ones through the computer simulations.