The Power of Localization for Efficiently Learning Linear Separators with Noise
This work addresses the challenge of robust machine learning for linear classifiers under adversarial conditions, offering incremental improvements in noise tolerance and efficiency for applications in secure or noisy data environments.
The paper tackles the problem of learning linear separators efficiently in the presence of noise, specifically in malicious and adversarial label noise models, by introducing a new localization-based approach that achieves noise tolerance rates of Ω(ε) and provides polynomial-time algorithms with polylogarithmic label complexity in active learning.
We introduce a new approach for designing computationally efficient learning algorithms that are tolerant to noise, and demonstrate its effectiveness by designing algorithms with improved noise tolerance guarantees for learning linear separators. We consider both the malicious noise model and the adversarial label noise model. For malicious noise, where the adversary can corrupt both the label and the features, we provide a polynomial-time algorithm for learning linear separators in $\Re^d$ under isotropic log-concave distributions that can tolerate a nearly information-theoretically optimal noise rate of $η= Ω(ε)$. For the adversarial label noise model, where the distribution over the feature vectors is unchanged, and the overall probability of a noisy label is constrained to be at most $η$, we also give a polynomial-time algorithm for learning linear separators in $\Re^d$ under isotropic log-concave distributions that can handle a noise rate of $η= Ω\left(ε\right)$. We show that, in the active learning model, our algorithms achieve a label complexity whose dependence on the error parameter $ε$ is polylogarithmic. This provides the first polynomial-time active learning algorithm for learning linear separators in the presence of malicious noise or adversarial label noise.