SutraNets: Sub-series Autoregressive Networks for Long-Sequence, Probabilistic Forecasting

TL;DR

SutraNets introduces a novel autoregressive approach that models long-sequence time series by decomposing them into lower-frequency sub-series, reducing error accumulation and improving forecasting accuracy.

Contribution

The paper presents SutraNets, a new neural probabilistic forecasting method that effectively models long sequences by leveraging sub-series decomposition to enhance accuracy and scalability.

Findings

Significantly outperforms existing methods on six real-world datasets.

Reduces error accumulation in long-sequence forecasting.

Maintains high accuracy when scaling model depth and width.

Abstract

We propose SutraNets, a novel method for neural probabilistic forecasting of long-sequence time series. SutraNets use an autoregressive generative model to factorize the likelihood of long sequences into products of conditional probabilities. When generating long sequences, most autoregressive approaches suffer from harmful error accumulation, as well as challenges in modeling long-distance dependencies. SutraNets treat long, univariate prediction as multivariate prediction over lower-frequency sub-series. Autoregression proceeds across time and across sub-series in order to ensure coherent multivariate (and, hence, high-frequency univariate) outputs. Since sub-series can be generated using fewer steps, SutraNets effectively reduce error accumulation and signal path distances. We find SutraNets to significantly improve forecasting accuracy over competitive alternatives on six real-world datasets, including when we vary the number of sub-series and scale up the depth and width of the underlying sequence models.