Members of the Euclid Team in collaboration with members of Archimedes Research Unit at Athena Research Center took part in the “NeurIPS 2023 - Machine Unlearning” competition held under the auspices of Google (part of NeurIPS 2023) by developing their own solution. The team competed and achieved 6th place out of 7 winners among 1120 submissions from 72 countries.

Problem Description

The competition considers a realistic scenario in which an age predictor has been trained on face images, and, after training, a certain subset of the training images must be forgotten to protect the privacy or rights of the individuals concerned. The goal was to submit code that takes as input the trained predictor, the forget and retain sets and outputs the weights of a predictor that has unlearned the designated forget set.

Participation

Our team developed a novel approach for machine unlearning using weight resetting and distillation techniques. The methodology and findings were highlighted in a presentantion by Google’s team at the NeurIPS 2023 Conference.