Recurrent Interpolants for Probabilistic Time Series Prediction

TL;DR

This paper introduces a novel probabilistic time series prediction method that combines recurrent neural networks with diffusion models, addressing scalability and dependency modeling challenges in high-dimensional data.

Contribution

It proposes a new approach integrating RNNs and diffusion models using stochastic interpolants for improved probabilistic forecasting.

Findings

Enhanced modeling of high-dimensional distributions

Improved scalability over existing generative approaches

Effective handling of cross-feature dependencies

Abstract

Sequential models like recurrent neural networks and transformers have become standard for probabilistic multivariate time series forecasting across various domains. Despite their strengths, they struggle with capturing high-dimensional distributions and cross-feature dependencies. Recent work explores generative approaches using diffusion or flow-based models, extending to time series imputation and forecasting. However, scalability remains a challenge. This work proposes a novel method combining recurrent neural networks' efficiency with diffusion models' probabilistic modeling, based on stochastic interpolants and conditional generation with control features, offering insights for future developments in this dynamic field.