Research Intern

Research Intern, April 2024 - Present

Vector Institute, Toronto, Canada

Project: Data-Free Machine Unlearning Algorithms

Supervisor: Dr. Rahul Krishnan

Description:

I work under the supervision of Dr. Rahul Krishnan at the Vector Institute continuing the work I started as an undergraduate researcher in his lab.

My research focuses on the development of machine unlearning algorithms for data removal. The goal of this project is to develop algorithms that can remove data from machine learning models in a way that is both efficient and effective. This research is motivated by the need to comply with privacy regulations such as the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA).

In this position, I am developing methods of unlearning that do not require access to the data being forgotten by exploiting gradient-basd weight saliency and sparsity in machine learning models. My work and project was entirely self-initiated and self-driven, and I have made significant progress in the field of privacy-preserving machine learning. With the help of Michael Cooper and Dr. Rahul Krishnan, I have been able to put my work into 2 sole first-author papers. The first is an application of my unlearning technique to the task of Tabular Representation Learning, and is currently under review at the Tabular Representation Learning Workshop at NeurIPS 2024. The second is a full paper on the unlearning algorithm I developed, an establishment of the field of blind unlearning, and an extension of unlearning into remedial learning for post-hoc fitting of trained models to arbitrary transformations to their training data. This paper is currently under review as a conference paper at The Thirteenth International Conference on Learning Representations (ICLR 2025).

This research is supported by a Vector Institute Research Grant.