A new back-propagation algorithm with momentum coefficient for medical datasets

Abstract

The standard backward propagation of errors algo rithm (abbreviated as back-propagation algorithm) is commonly used for decision making in medicine. Using the back-propagation algorithm in medical diagnosis is desirable since it avoids human sub jectivity and applies a large knowledge base, which makes this algorithm very reliable. However, it is generally believed that it is very slow if it does con verge, especially if the network size is not of suf ficient size compared to the problem at hand. A drawback of the back-propagation algorithm is that it has a constant learning rate coefficient, while dif ferent regions of the error surface may have differ ent characteristic gradients. Variation in the nature of the surface may require a dynamic change of learning rate coefficient. A new back-propagation algorithm with momentum has been developed in order to be used to speed up the learning pro cess, which accelerates the convergence of back propagation algorithm.