Targeted Unlearning Using Perturbed Sign Gradient Methods with Applications on Medical Images

Abstract

Machine unlearning aims to remove the influence of specific training samples from a trained model without full retraining. While prior work has largely focused on privacy-motivated settings, we recast unlearning as a general-purpose tool for post-deployment model revision. Specifically, we focus on utilizing unlearning in clinical contexts where data shifts, device deprecation, and policy changes are common. To this end, we propose a bilevel optimization formulation of boundary-based unlearning that can be solved using iterative algorithms. We provide convergence guarantees when first order algorithms are used to unlearn and introduce a tunable loss design for controlling the forgetting–retention tradeoff. Across benchmark and real-world clinical imaging datasets, our approach outperforms baselines on both forgetting and retention metrics, including scenarios involving imaging devices and anatomical outliers. This work demonstrates the feasibility of unlearning on clinical imaging datasets and proposes it as a tool for model maintenance in scenarios that require removing the influence of specific data points without full model retraining. Code is available $\href{https://github.com/monkeygobah/unlearning_langevin}{here}$.