Differentially Private Continual Learning

Sebastian Farquhar; Yarin Gal

Differentially Private Continual Learning

Machine Learning 2019-02-19 v1 Machine Learning

Authors: Sebastian Farquhar , Yarin Gal

Abstract

Catastrophic forgetting can be a significant problem for institutions that must delete historic data for privacy reasons. For example, hospitals might not be able to retain patient data permanently. But neural networks trained on recent data alone will tend to forget lessons learned on old data. We present a differentially private continual learning framework based on variational inference. We estimate the likelihood of past data given the current model using differentially private generative models of old datasets.

Keywords

differential privacy deep learning multi-task learning

Cite

@article{arxiv.1902.06497,
  title  = {Differentially Private Continual Learning},
  author = {Sebastian Farquhar and Yarin Gal},
  journal= {arXiv preprint arXiv:1902.06497},
  year   = {2019}
}

Comments

Presented at the Privacy in Machine Learning and AI workshop at ICML 2018

Related papers

View all related →