Title: Controllable Machine Unlearning via Gradient Pivoting Show Chinese title

Title: 通过梯度旋转的可控机器遗忘

Abstract: Machine unlearning (MU) aims to remove the influence of specific data from a trained model. However, approximate unlearning methods, often formulated as a single-objective optimization (SOO) problem, face a critical trade-off between unlearning efficacy and model fidelity. This leads to three primary challenges: the risk of over-forgetting, a lack of fine-grained control over the unlearning process, and the absence of metrics to holistically evaluate the trade-off. To address these issues, we reframe MU as a multi-objective optimization (MOO) problem. We then introduce a novel algorithm, Controllable Unlearning by Pivoting Gradient (CUP), which features a unique pivoting mechanism. Unlike traditional MOO methods that converge to a single solution, CUP's mechanism is designed to controllably navigate the entire Pareto frontier. This navigation is governed by a single intuitive hyperparameter, the `unlearning intensity', which allows for precise selection of a desired trade-off. To evaluate this capability, we adopt the hypervolume indicator, a metric that captures both the quality and diversity of the entire set of solutions an algorithm can generate. Our experimental results demonstrate that CUP produces a superior set of Pareto-optimal solutions, consistently outperforming existing methods across various vision tasks. Show Chinese abstract

Abstract: 机器遗忘（MU）旨在从训练好的模型中移除特定数据的影响。 然而，近似遗忘方法通常被表述为单目标优化（SOO）问题，面临着遗忘效果和模型保真度之间的关键权衡。 这导致了三个主要挑战：过度遗忘的风险、对遗忘过程缺乏细粒度控制以及缺乏全面评估权衡的指标。 为了解决这些问题，我们将MU重新表述为多目标优化（MOO）问题。 然后，我们引入了一种新的算法，通过枢轴梯度进行可控遗忘（CUP），该算法具有独特的枢轴机制。 与传统MOO方法收敛到单一解不同，CUP的机制旨在可控地遍历整个帕累托前沿。 这种遍历由一个直观的超参数“遗忘强度”控制，允许精确选择所需的权衡。 为了评估这种能力，我们采用了超体积指标，该指标能够捕捉算法可以生成的所有解的质量和多样性。 我们的实验结果表明，CUP生成了一组优越的帕累托最优解，在各种视觉任务中始终优于现有方法。