Autoregressive Convolutional Neural Networks for Asynchronous Time Series

TL;DR

This paper introduces Significance-Offset Convolutional Neural Networks, a novel deep learning architecture for regression tasks on asynchronous multivariate time series, inspired by autoregressive models and gating mechanisms.

Contribution

The paper presents a new convolutional neural network architecture tailored for asynchronous time series, combining autoregressive weighting with data-dependent convolutional functions, demonstrating improved performance.

Findings

Achieves promising results on diverse asynchronous datasets.

Outperforms standard convolutional and recurrent neural networks.

Effective for large-scale financial and household electricity data.

Abstract

We propose Significance-Offset Convolutional Neural Network, a deep convolutional network architecture for regression of multivariate asynchronous time series. The model is inspired by standard autoregressive (AR) models and gating mechanisms used in recurrent neural networks. It involves an AR-like weighting system, where the final predictor is obtained as a weighted sum of adjusted regressors, while the weights are datadependent functions learnt through a convolutional network. The architecture was designed for applications on asynchronous time series and is evaluated on such datasets: a hedge fund proprietary dataset of over 2 million quotes for a credit derivative index, an artificially generated noisy autoregressive series and UCI household electricity consumption dataset. The proposed architecture achieves promising results as compared to convolutional and recurrent neural networks.