A method for decreasing neural network training time as applied to ECG classification

The single-layer feedforward neural network (FFNN) in conjunction with the backpropagation training algorithm (BPTA) is used for electrocardiogram (ECG) classification. It has been observed that, for such a problem, the values of the input weights are closely related to the input training set. An implication of this observation is that, rather than choosing initially random weights for the BPTA, one may choose initial weights that are actually quite close to an optimal solution. An advantage of such a choice is faster convergence time based on knowledge of the incoming training data. Decreasing convergence time makes more promising the use of the FFNN to classify ECGs for arrythmia detection, ambulatory monitoring and analysis, and front-line physician support instrumentation.<<ETX>>