MUNBa: Machine Unlearning via Nash Bargaining

TL;DR

MUNBa introduces a Nash bargaining-based approach to machine unlearning, effectively balancing forgetting specific data and preserving model utility, with proven improvements across diverse vision tasks.

Contribution

The paper formulates machine unlearning as a Nash bargaining game, providing a novel closed-form solution that guarantees an equilibrium for optimal forgetting and preservation.

Findings

Outperforms state-of-the-art MU algorithms in various tasks

Improves forgetting precision and model robustness

Enhances trade-off between forgetting and preservation

Abstract

Machine Unlearning (MU) aims to selectively erase harmful behaviors from models while retaining the overall utility of the model. As a multi-task learning problem, MU involves balancing objectives related to forgetting specific concepts/data and preserving general performance. A naive integration of these forgetting and preserving objectives can lead to gradient conflicts and dominance, impeding MU algorithms from reaching optimal solutions. To address the gradient conflict and dominance issue, we reformulate MU as a two-player cooperative game, where the two players, namely, the forgetting player and the preservation player, contribute via their gradient proposals to maximize their overall gain and balance their contributions. To this end, inspired by the Nash bargaining theory, we derive a closed-form solution to guide the model toward the Pareto stationary point. Our formulation of MU guarantees an equilibrium solution, where any deviation from the final state would lead to a reduction in the overall objectives for both players, ensuring optimality in each objective. We evaluate our algorithm's effectiveness on a diverse set of tasks across image classification and image generation. Extensive experiments with ResNet, vision-language model CLIP, and text-to-image diffusion models demonstrate that our method outperforms state-of-the-art MU algorithms, achieving a better trade-off between forgetting and preserving. Our results also highlight improvements in forgetting precision, preservation of generalization, and robustness against adversarial attacks.