On Optimal Learning Under Targeted Data Poisoning

• URL: http://arxiv.org/abs/2210.02713v1
• Date: Thu, 6 Oct 2022 06:49:48 GMT
• Title: On Optimal Learning Under Targeted Data Poisoning
• Authors: Steve Hanneke, Amin Karbasi, Mohammad Mahmoody, Idan Mehalel and Shay Moran
• Abstract summary: In this work we aim to characterize the smallest achievable error $epsilon=epsilon(eta)$ by the learner in the presence of such an adversary. Remarkably, we show that the upper bound can be attained by a deterministic learner.
• Score: 48.907813854832206
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Consider the task of learning a hypothesis class $\mathcal{H}$ in the presence of an adversary that can replace up to an $\eta$ fraction of the examples in the training set with arbitrary adversarial examples. The adversary aims to fail the learner on a particular target test point $x$ which is known to the adversary but not to the learner. In this work we aim to characterize the smallest achievable error $\epsilon=\epsilon(\eta)$ by the learner in the presence of such an adversary in both realizable and agnostic settings. We fully achieve this in the realizable setting, proving that $\epsilon=\Theta(\mathtt{VC}(\mathcal{H})\cdot \eta)$, where $\mathtt{VC}(\mathcal{H})$ is the VC dimension of $\mathcal{H}$. Remarkably, we show that the upper bound can be attained by a deterministic learner. In the agnostic setting we reveal a more elaborate landscape: we devise a deterministic learner with a multiplicative regret guarantee of $\epsilon \leq C\cdot\mathtt{OPT} + O(\mathtt{VC}(\mathcal{H})\cdot \eta)$, where $C > 1$ is a universal numerical constant. We complement this by showing that for any deterministic learner there is an attack which worsens its error to at least $2\cdot \mathtt{OPT}$. This implies that a multiplicative deterioration in the regret is unavoidable in this case. Finally, the algorithms we develop for achieving the optimal rates are inherently improper. Nevertheless, we show that for a variety of natural concept classes, such as linear classifiers, it is possible to retain the dependence $\epsilon=\Theta_{\mathcal{H}}(\eta)$ by a proper algorithm in the realizable setting. Here $\Theta_{\mathcal{H}}$ conceals a polynomial dependence on $\mathtt{VC}(\mathcal{H})$.

Related papers

• Deterministic Apple Tasting [2.4554686192257424]
  We provide the first widely-applicable deterministic apple tasting learner. We prove a trichotomy stating that every class $mathcalH$ must be either easy, hard, or unlearnable. Our upper bound is based on a deterministic algorithm for learning from expert advice with apple tasting feedback.
  arXiv  Detail & Related papers  (2024-10-14T11:54:46Z)
• Revisiting Agnostic PAC Learning [30.67561230812141]
  PAC learning, dating back to Valiant'84 and Vapnik and Chervonenkis'64,'74, is a classic model for studying supervised learning. Empirical Risk Minimization (ERM) is a natural learning algorithm, where one simply outputs the hypothesis from $mathcalH$ making the fewest mistakes on the training data. We revisit PAC learning and first show that ERM is in fact sub-optimal if we treat the performance of the best hypothesis, denoted $tau:=Pr_mathcalD[hstar_math
  arXiv  Detail & Related papers  (2024-07-29T08:20:49Z)
• Online Learning of Halfspaces with Massart Noise [47.71073318490341]
  We study the task of online learning in the presence of Massart noise. We present a computationally efficient algorithm that achieves mistake bound $eta T + o(T)$. We use our Massart online learner to design an efficient bandit algorithm that obtains expected reward at least $(1-1/k) Delta T - o(T)$ bigger than choosing a random action at every round.
  arXiv  Detail & Related papers  (2024-05-21T17:31:10Z)
• Cryptographic Hardness of Learning Halfspaces with Massart Noise [59.8587499110224]
  We study the complexity of PAC learning halfspaces in the presence of Massart noise. We show that no-time Massart halfspace learners can achieve error better than $Omega(eta)$, even if the optimal 0-1 error is small.
  arXiv  Detail & Related papers  (2022-07-28T17:50:53Z)
• Agnostic learning with unknown utilities [70.14742836006042]
  In many real-world problems, the utility of a decision depends on the underlying context $x$ and decision $y$. We study this as agnostic learning with unknown utilities. We show that estimating the utilities of only the sampled points$S$ suffices to learn a decision function which generalizes well.
  arXiv  Detail & Related papers  (2021-04-17T08:22:04Z)
• Optimal Regret Algorithm for Pseudo-1d Bandit Convex Optimization [51.23789922123412]
  We study online learning with bandit feedback (i.e. learner has access to only zeroth-order oracle) where cost/reward functions admit a "pseudo-1d" structure. We show a lower bound of $min(sqrtdT, T3/4)$ for the regret of any algorithm, where $T$ is the number of rounds. We propose a new algorithm sbcalg that combines randomized online gradient descent with a kernelized exponential weights method to exploit the pseudo-1d structure effectively.
  arXiv  Detail & Related papers  (2021-02-15T08:16:51Z)
• Phase Transitions in Rate Distortion Theory and Deep Learning [5.145741425164946]
  We say that $mathcalS$ can be compressed at rate $s$ if we can achieve an error of $mathcalO(R-s)$ for encoding $mathcalS$. We show that for certain "nice" signal classes $mathcalS$, a phase transition occurs: We construct a probability measure $mathbbP$ on $mathcalS$.
  arXiv  Detail & Related papers  (2020-08-03T16:48:49Z)
• Taking a hint: How to leverage loss predictors in contextual bandits? [63.546913998407405]
  We study learning in contextual bandits with the help of loss predictors. We show that the optimal regret is $mathcalO(minsqrtT, sqrtmathcalETfrac13)$ when $mathcalE$ is known.
  arXiv  Detail & Related papers  (2020-03-04T07:36:38Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.