Contrastive Time Series Forecasting with Anomalies

Name: Contrastive Time Series Forecasting with Anomalies
Uploaded: 2026-10-06T07:21:19.993884+00:00
Duration: 682 s
Description: Co-TSFA is a regularization method that helps time-series models ignore irrelevant noise while adapting to important forecast anomalies. It uses contrastive learning to align internal representations with actual output changes, ensuring the model only adjusts its predictions when the underlying data shift is meaningful for forecasting accuracy.

Co-TSFA is a regularization method that helps time-series models ignore irrelevant noise while adapting to important forecast anomalies. It uses contrastive learning to align internal representations with actual output changes, ensuring the model only adjusts its predictions when the underlying data shift is meaningful for forecasting accuracy.

Open the full episode