SCINet: Time Series Modeling and Forecasting with Sample Convolution and Interaction

TL;DR

SCINet is a novel neural network architecture that leverages sample convolution and interaction to effectively model complex temporal dynamics in time series forecasting, outperforming existing methods.

Contribution

Introduces SCINet, a recursive downsample-convolve-interact neural network architecture that captures multi-resolution temporal features for improved forecasting accuracy.

Findings

SCINet outperforms existing convolutional and Transformer models on multiple datasets.

SCINet effectively models complex temporal dependencies in time series.

The architecture demonstrates significant improvements in forecasting accuracy.

Abstract

One unique property of time series is that the temporal relations are largely preserved after downsampling into two sub-sequences. By taking advantage of this property, we propose a novel neural network architecture that conducts sample convolution and interaction for temporal modeling and forecasting, named SCINet. Specifically, SCINet is a recursive downsample-convolve-interact architecture. In each layer, we use multiple convolutional filters to extract distinct yet valuable temporal features from the downsampled sub-sequences or features. By combining these rich features aggregated from multiple resolutions, SCINet effectively models time series with complex temporal dynamics. Experimental results show that SCINet achieves significant forecasting accuracy improvements over both existing convolutional models and Transformer-based solutions across various real-world time series forecasting datasets. Our codes and data are available at https://github.com/cure-lab/SCINet.