We study the problem of PAC learning halfspaces on $\mathbb{R}^d$ with Massart noise under the Gaussian distribution. In the Massart model, an adversary is allowed to flip the label of each point $\mathbf{x}$ with unknown probability $\eta(\mathbf{x}) \leq \eta$, for some parameter $\eta \in [0,1/2]$. The goal is to find a hypothesis with misclassification error of $\mathrm{OPT} + \epsilon$, where $\mathrm{OPT}$ is the error of the target halfspace. This problem had been previously studied under two assumptions: (i) the target halfspace is homogeneous (i.e., the separating hyperplane goes through the origin), and (ii) the parameter $\eta$ is strictly smaller than $1/2$. Prior to this work, no nontrivial bounds were known when either of these assumptions is removed. We study the general problem and establish the following: For $\eta <1/2$, we give a learning algorithm for general halfspaces with sample and computational complexity $d^{O_{\eta}(\log(1/\gamma))}\mathrm{poly}(1/\epsilon)$, where $\gamma =\max\{\epsilon, \min\{\mathbf{Pr}[f(\mathbf{x}) = 1], \mathbf{Pr}[f(\mathbf{x}) = -1]\} \}$ is the bias of the target halfspace $f$. Prior efficient algorithms could only handle the special case of $\gamma = 1/2$. Interestingly, we establish a qualitatively matching lower bound of $d^{\Omega(\log(1/\gamma))}$ on the complexity of any Statistical Query (SQ) algorithm. For $\eta = 1/2$, we give a learning algorithm for general halfspaces with sample and computational complexity $O_\epsilon(1) d^{O(\log(1/\epsilon))}$. This result is new even for the subclass of homogeneous halfspaces; prior algorithms for homogeneous Massart halfspaces provide vacuous guarantees for $\eta=1/2$. We complement our upper bound with a nearly-matching SQ lower bound of $d^{\Omega(\log(1/\epsilon))}$, which holds even for the special case of homogeneous halfspaces.
翻译:我们研究 PAC 在 $\ mathb{R{% d$ 上学习半空的问题。 目标是在 Gaus 分布下找到一个错误分解错误的假设 $\ mathb{R{ =d$; 在 Massart 模型中, 允许对手翻转每个点的标签$\ mathbf{x} 美元, 概率未知 $eta (mathb{x}x} ) 问题。 对于某些参数 $\eta 美元 [0/ 2] 美元。 目标是找到一个错误错误的假设 $\ mathrm{ { {OPT} +\ epsal$, 其中$\ mass 美元是目标半空间的错误 $\gx} 。 问题先前在两个假设下研究 : (i) 目标半空间的分解结果在源头, 参数 $\ a 0. 2美元 。 在这项工作之前, 当这些假设被删除时, 我们研究一般问题, 并设定如下: $\\\\\\\\\\\\\ ma\\\\ ma= dal adal rialalalal max exal max.