Capsule Neural Networks as Noise Stabilizer for Time Series Data

arXiv:2403.13867v1 Announce Type: new
Abstract: Capsule Neural Networks utilize capsules, which bind neurons into a single vector and learn position equivariant features, which makes them more robust than original Convolutional Neural Networks. CapsNets employ an affine transformation matrix and dynamic routing with coupling coefficients to learn robustly. In this paper, we investigate the effectiveness of CapsNets in analyzing highly sensitive and noisy time series sensor data. To demonstrate CapsNets robustness, we compare their performance with original CNNs on electrocardiogram data, …

abstract arxiv capsule convolutional neural networks cs.lg data dynamic features learn matrix networks neural networks neurons noise paper robust routing series them time series transformation type vector