Decoupled self-supervised deep multi-task learning framework for subscriber portrait in smart meter

doi:10.1016/j.patcog.2025.112633

Decoupled self-supervised deep multi-task learning framework for subscriber portrait in smart meter

2025-10-27

0

PRE

AI

DOI：10.1016/j.patcog.2025.112633

Original

Original request for help

Save

Abstract

En 中文

• A decoupled self-supervised deep learning framework for subscriber portraits is proposed named DSS-MTL. • Anovel data augmentation method based on information entropy for smart meter data is proposed, enhancing encoder performance. • A method for quantifying task correlation has been developed, guiding the training process of deep multi-task learning. • The proposed DSS-MTL has achieved state-of-the-art performance in both intra-domain and inter-domain tasks for subscriber portraits based on smart meter data.

Decoupled self-supervised deep multi-task learning framework for subscriber portrait in smart meter

Abstract

Journal

Organization