Gradient Surgery for One-shot Unlearning on Generative Model

TL;DR

This paper introduces a gradient manipulation technique inspired by multi-task learning to efficiently unlearn specific data influence in deep generative models, outperforming existing methods and providing theoretical insights.

Contribution

Proposes a novel gradient projection method for unlearning in generative models, with theoretical analysis and superior empirical performance.

Findings

Outperforms existing unlearning baselines.

Provides theoretical analysis of the unlearning process.

Effective in removing data influence from generative models.

Abstract

Recent regulation on right-to-be-forgotten emerges tons of interest in unlearning pre-trained machine learning models. While approximating a straightforward yet expensive approach of retrain-from-scratch, recent machine unlearning methods unlearn a sample by updating weights to remove its influence on the weight parameters. In this paper, we introduce a simple yet effective approach to remove a data influence on the deep generative model. Inspired by works in multi-task learning, we propose to manipulate gradients to regularize the interplay of influence among samples by projecting gradients onto the normal plane of the gradients to be retained. Our work is agnostic to statistics of the removal samples, outperforming existing baselines while providing theoretical analysis for the first time in unlearning a generative model.