Exogenous Dropout Enhances Time Series Forecasting Robustness

2026-07-08

A novel method, exogenous dropout, has been introduced to improve the resilience of time series forecasting models that utilize exogenous covariates. The technique demonstrates significant gains in robustness against noisy, misaligned, or missing covariate data.

Source: arXiv · cs.LG

Reported by VERA Newswire.