Neural networks for ECG compression and classification
I H Habboush, George B Moody, R.G. Mark · 2002
The authors compared neural networks designed for electrocardiogram (ECG) compression and classification with optimum linear methods. It is found that simple neural networks with one hidden layer approach the performance of linear methods, but offer no advantage over them. Suitably constructed networks with more than one hidden layer, however, can perform more efficient ECG compression than is possible using linear methods under the same constraints.>