A Characterization of Semi-Supervised Adversarially-Robust PAC Learnability
share
We study the problem of semi-supervised learning of an adversarially-robust predictor in the PAC model, where the learner has access to both labeled and unlabeled examples. The sample complexity in semi-supervised learning has two parameters, the number of labeled examples and the number of unlabeled examples. We consider the complexity measures, VC_U ≤ dim_U ≤ VC and VC^*, where VC is the standard VC-dimension, VC^* is its dual, and the other two measures appeared in Montasser et al. (2019). The best sample bound known for robust supervised PAC learning is O(VC · VC^*), and we will compare our sample bounds to Λ which is the minimal number of labeled examples required by any robust supervised PAC learning algorithm. Our main results are the following: (1) in the realizable setting it is sufficient to have O(VC_U) labeled examples and O(Λ) unlabeled examples. (2) In the agnostic setting, let η be the minimal agnostic error. The sample complexity depends on the resulting error rate. If we allow an error of 2η+ϵ, it is still sufficient to have O(VC_U) labeled examples and O(Λ) unlabeled examples. If we insist on having an error η+ϵ then Ω(dim_U) labeled examples are necessary, as in the supervised case. The above results show that there is a significant benefit in semi-supervised robust learning, as there are hypothesis classes with VC_U=0 and dim_U arbitrary large. In supervised learning, having access only to labeled examples requires at least Λ≥ dim_U labeled examples. Semi-supervised require only O(1) labeled examples and O(Λ) unlabeled examples. A byproduct of our result is that if we assume that the distribution is robustly realizable by a hypothesis class, then with respect to the 0-1 loss we can learn with only O(VC_U) labeled examples, even if the VC is infinite.
READ FULL TEXT
