A homotopy method for training neural networks

Development of a calibrating algorithm for Delta Robot’s visual positioning based on artificial neural network

“…Scholars have studied on the learning process of artificial neural network: a) The improvement of error function and adaptive adjustment of learning rate (such as method of entropy function [14], conjugate gradient method [15] and simplex method [16] ; b) The use of optimization algorithm and parameter estimation theory in other field such as Kalman Filtering [17], Homotopy theory [18] and orthogonal projection algorithm [19]. There are still other improved algorithms which are designed specifically to overcome local minimum.…”