Learning dynamic representations via an optimally-weighted maximum mean discrepancy optimization framework for continual learning

Kaihui Huang

doi:10.1016/j.knosys.2026.115419

Learning dynamic representations via an optimally-weighted maximum mean discrepancy optimization framework for continual learning

2026-01-30

0

PRE

AI

DOI：10.1016/j.knosys.2026.115419

Original

Original request for help

Save

Abstract

En 中文

• An innovative framework termed Optimally Weighted Maximum Mean Discrepancy (OWMMD) is proposed to mitigate catastrophic forgetting in continual learning paradigms. • A Multi-Level Feature Matching Mechanism (MLFMM) is pro impose penalties on the modification of feature representations across various tasks. • An Adaptive Regularization Optimization (ARO) framework that enables the model to evaluate the significance of each feature layer in real-time throughout the optimization process.

Keywords:

Continual Learning

Catastrophic Forgetting

Maximum Mean Discrepancy

Feature Matching

Adaptive Regularization

Learning dynamic representations via an optimally-weighted maximum mean discrepancy optimization framework for continual learning

Abstract

Journal

Organization