English

From Adaptive Query Release to Machine Unlearning

Machine Learning 2023-07-24 v1 Machine Learning

Abstract

We formalize the problem of machine unlearning as design of efficient unlearning algorithms corresponding to learning algorithms which perform a selection of adaptive queries from structured query classes. We give efficient unlearning algorithms for linear and prefix-sum query classes. As applications, we show that unlearning in many problems, in particular, stochastic convex optimization (SCO), can be reduced to the above, yielding improved guarantees for the problem. In particular, for smooth Lipschitz losses and any ρ>0\rho>0, our results yield an unlearning algorithm with excess population risk of O~(1n+dnρ)\tilde O\big(\frac{1}{\sqrt{n}}+\frac{\sqrt{d}}{n\rho}\big) with unlearning query (gradient) complexity O~(ρRetraining Complexity)\tilde O(\rho \cdot \text{Retraining Complexity}), where dd is the model dimensionality and nn is the initial number of samples. For non-smooth Lipschitz losses, we give an unlearning algorithm with excess population risk O~(1n+(dnρ)1/2)\tilde O\big(\frac{1}{\sqrt{n}}+\big(\frac{\sqrt{d}}{n\rho}\big)^{1/2}\big) with the same unlearning query (gradient) complexity. Furthermore, in the special case of Generalized Linear Models (GLMs), such as those in linear and logistic regression, we get dimension-independent rates of O~(1n+1(nρ)2/3)\tilde O\big(\frac{1}{\sqrt{n}} +\frac{1}{(n\rho)^{2/3}}\big) and O~(1n+1(nρ)1/3)\tilde O\big(\frac{1}{\sqrt{n}} +\frac{1}{(n\rho)^{1/3}}\big) for smooth Lipschitz and non-smooth Lipschitz losses respectively. Finally, we give generalizations of the above from one unlearning request to \textit{dynamic} streams consisting of insertions and deletions.

Keywords

Cite

@article{arxiv.2307.11228,
  title  = {From Adaptive Query Release to Machine Unlearning},
  author = {Enayat Ullah and Raman Arora},
  journal= {arXiv preprint arXiv:2307.11228},
  year   = {2023}
}

Comments

Accepted to ICML 2023

R2 v1 2026-06-28T11:36:28.563Z