Convolutional signature for sequential data

Abstract Signature is an infinite graded sequence of statistics known to characterize geometric rough paths. While the use signature in machine learning successful low-dimensional cases, it suffers from curse dimensionality high-dimensional as number features truncated transform grows exponentially fast. With idea Convolutional Neural Network, we propose a novel neural network address this problem. Our model reduces efficiently data-dependent way. Some empirical experiments including financial time series classification and natural language processing are provided support our convolutional model.