TRAJECTORY METHODS FOR SUPERVISED LEARNING

The task of supervised learning in Artificial Neural Networks reduces to the minimization of the network error function subject to the weights of the network. Trajectory Methods are global optimization methods that formulate the optimization problem into a set of ordinary differential equations, the equilibrium points of which, correspond to local minima of the objective function. In this work, we apply Trajectory methods to address the Neural Network training problem. The reported experimental results indicate that this is a promising approach.