Semantic Residual Prompts for Continual Learning: A Novel Approach to Enhance Stability and Plasticity
Stärkung der Stabilität und Plastizität durch zweistufige Prompt-Strategie.
GUIDE: Guidance-based Incremental Learning with Diffusion Models
GUIDE introduces a novel approach that utilizes classifier guidance to generate rehearsal samples targeting forgotten information, reducing catastrophic forgetting in continual learning.
AttriCLIP: A Non-Incremental Learner for Incremental Knowledge Learning
AttriCLIP is a non-incremental learner that incrementally extracts knowledge of new classes or tasks without increasing model parameters or requiring replay data.
AttriCLIP: A Non-Incremental Learner for Incremental Knowledge Learning
AttriCLIP is a non-incremental learner that incrementally extracts knowledge of new classes or tasks without the need for additional memory, outperforming previous state-of-the-art methods in realistic settings.
Ensemble of Selectively Trained Experts in Continual Learning: SEED Method
SEED method introduces a novel approach to continual learning by selectively training experts, mitigating forgetting, encouraging diversification, and maintaining high plasticity.
GUIDE: Guidance-based Incremental Learning with Diffusion Models
新しい連続学習手法の提案とその効果的な実装に焦点を当てる。
Semantic Residual Prompts for Continual Learning: A Novel Approach to Enhance Stability and Plasticity in CL
The author proposes STAR-Prompt, a two-level prompting strategy that leverages a foundation model to enhance stability and plasticity in Continual Learning. By introducing semantic residuals and generative replay, the method outperforms existing approaches.
GUIDE: Guidance-based Incremental Learning with Diffusion Models
The author introduces GUIDE, a novel continual learning approach that utilizes diffusion models to generate rehearsal samples targeting forgotten information. By integrating classifier guidance techniques, the method reduces catastrophic forgetting and outperforms conventional generative replay methods.
