Winner-takes-all for Multivariate Probabilistic Time Series Forecasting

TL;DR

This paper presents TimeMCL, a neural network method using Multiple Choice Learning and Winner-Takes-All loss to generate diverse, plausible future predictions for multivariate time series, addressing ambiguity efficiently.

Contribution

It adapts Multiple Choice Learning with WTA loss for time series forecasting, enabling diverse predictions with low computational cost.

Findings

Demonstrates promising performance on real-world data.

Provides insights through synthetic data experiments.

Achieves diversity in predictions effectively.

Abstract

We introduce TimeMCL, a method leveraging the Multiple Choice Learning (MCL) paradigm to forecast multiple plausible time series futures. Our approach employs a neural network with multiple heads and utilizes the Winner-Takes-All (WTA) loss to promote diversity among predictions. MCL has recently gained attention due to its simplicity and ability to address ill-posed and ambiguous tasks. We propose an adaptation of this framework for time-series forecasting, presenting it as an efficient method to predict diverse futures, which we relate to its implicit quantization objective. We provide insights into our approach using synthetic data and evaluate it on real-world time series, demonstrating its promising performance at a light computational cost.