Go Beyond Multiple Instance Neural Networks: Deep-learning Models based on Local Pattern Aggregation
This work addresses a practical issue in clinical ECG analysis and other domains where input sizes vary, offering an incremental improvement over existing methods.
The paper tackles the problem of processing variable-size data with deep learning by proposing LPANet, a model based on local pattern aggregation, which reduces parameter tuning difficulty and improves generalization, showing advantages over CNN and LSTM in premature ventricular contraction detection.
Deep convolutional neural networks (CNNs) have brought breakthroughs in processing clinical electrocardiograms (ECGs), speaker-independent speech and complex images. However, typical CNNs require a fixed input size while it is common to process variable-size data in practical use. Recurrent networks such as long short-term memory (LSTM) are capable of eliminating the restriction, but suffer from high computational complexity. In this paper, we propose local pattern aggregation-based deep-learning models to effectively deal with both problems. The novel network structure, called LPANet, has cropping and aggregation operations embedded into it. With these new features, LPANet can reduce the difficulty of tuning model parameters and thus tend to improve generalization performance. To demonstrate the effectiveness, we applied it to the problem of premature ventricular contraction detection and the experimental results shows that our proposed method has certain advantages compared to classical network models, such as CNN and LSTM.