Trajectory Methods for Neural Network Training

Y.G. Petalas, D.K. Tasoulis, and M.N. Vrahatis (Greece)


Neural Networks, Training Algorithms, Ordinary Differen tial Equations, Trajectory Methods


A new class of methods for training multilayer feedfor ward neural networks is proposed. The proposed class of methods draws from methods for solving initial value prob lems of ordinary differential equations, and belong to the subclass of trajectory methods. The training of a multi layer feedforward neural network is equivalent to the mini mization of the network's error function with respect to the weights of the network. To address this problem we solve the differential equation "!$#&%')( , where is the vec tor of network weights and !$# is the gradient of the error function of the network. The solution of the above system of ordinary differential equations corresponds to the solu tion of the aforementioned minimization problem.

Important Links:

Go Back