DUCK: Distance-based Unlearning via Centroid Kinematics

TL;DR

DUCK is a novel machine unlearning algorithm that effectively removes specific data influences from neural models using centroid kinematics and metric learning, achieving state-of-the-art results and providing new evaluation metrics.

Contribution

The paper introduces DUCK, a new unlearning method leveraging centroid kinematics and metric learning, along with a novel evaluation metric, AUS, for assessing unlearning effectiveness.

Findings

DUCK achieves state-of-the-art unlearning performance on benchmark datasets.

AUS effectively measures unlearning success and model performance loss.

Analysis reveals how DUCK impacts feature space organization.

Abstract

Machine Unlearning is rising as a new field, driven by the pressing necessity of ensuring privacy in modern artificial intelligence models. This technique primarily aims to eradicate any residual influence of a specific subset of data from the knowledge acquired by a neural model during its training. This work introduces a novel unlearning algorithm, denoted as Distance-based Unlearning via Centroid Kinematics (DUCK), which employs metric learning to guide the removal of samples matching the nearest incorrect centroid in the embedding space. Evaluation of the algorithm's performance is conducted across various benchmark datasets in two distinct scenarios, class removal, and homogeneous sampling removal, obtaining state-of-the-art performance. We also introduce a novel metric, called Adaptive Unlearning Score (AUS), encompassing not only the efficacy of the unlearning process in forgetting target data but also quantifying the performance loss relative to the original model. Additionally, we conducted a thorough investigation of the unlearning mechanism in DUCK, examining its impact on the organization of the feature space and employing explainable AI techniques for deeper insights.