Regularized B-spline Network and its Application to Heart Arrhythmia Classification

摘要

This paper presents an effective learning scheme that combines B-spline modeling and regularized neural networks. Essential issues of structural design and learning process are discussed. Regularization theory is leveraged to design the topological structure of the network. A training algorithm is derived for the learning of both synaptic weights and B-spline coefficients. The approach is then applied to the medical problem of heart arrhythmia detection, particularly the detection of premature ventricular contraction. Promising results demonstrate the potential benefits of the proposed method.