Search papers, labs, and topics across Lattice.
This paper introduces the Trajectory-guided Forget-Recover Network (TFR-Net), a novel approach to continual unlearning in machine learning models that addresses the challenges of knowledge re-emergence and capacity reduction during repeated unlearning interventions. By tracking channel-level risk, TFR-Net effectively differentiates between persistent and transient target-related channels, suppressing only those that are persistent while reactivating dormant channels to recover model capacity. Experimental results across four datasets demonstrate that TFR-Net achieves a superior balance between unlearning effectiveness and retained utility compared to existing methods.
TFR-Net not only prevents knowledge re-emergence during continual unlearning but also enhances model capacity, achieving a remarkable trade-off between unlearning and utility retention.
Machine unlearning aims to eliminate the influence of sensitive data on a model. In the real world, unlearning requests arrive continually, which gives rise to two challenges. First, an unlearning intervention may redistribute target-related computation across remaining pathways, allowing previously forgotten knowledge to re-emerge. Second, repeated unlearning interventions may progressively reduce the model capacity needed to preserve retained utility. To address these challenges, we propose the Trajectory-guided Forget-Recover Network (TFR-Net). TFR-Net tracks channel-level risk across requests. It separates persistent target-related channels from transient hotspots and suppresses only the persistent ones. TFR-Net also recovers model capacity by reactivating dormant channels. These channels make strong contributions to retained utility and show low current and historical forget risk. The recovery is accepted only when retained-utility degradation remains within a predefined tolerance. Experiments on four datasets show that TFR-Net consistently achieves a more favorable trade-off between unlearning effectiveness and retained utility than representative baselines.
