Randomized hypotheses and minimum disagreement hypotheses for learning with noise

Abstract

In this paper we prove various results about PAC learning in the presence of malicious and random classification noise. Our main theme is the use of randomized hypotheses for learning with small sample sizes and high malicious noise rates. We show an algorithm that PAC learns any target class of VC-dimension d using randomized hypotheses and order of d/ε training examples (up to logarithmic factors) while tolerating malicious noise rates even slightly larger than the information-theoretic bound ε/(1+ε) for deterministic hypotheses. Combined with previous results, this implies that a lower bound d/Δ+ε/Δ ² on the sample size, where η=ε/(l+ε)−Δ is the malicious noise rate, applies only when using deterministic hypotheses. We then show that the information-theoretic upper bound on the noise rate for deterministic hypotheses can be replaced by 2ε/(l+2ε) if randomized hypotheses are used. Investigating further the use of randomized hypotheses, we show a strategy for learning the powerset of d elements using an optimal sample size of order dε/Δ ² (up to logarithmic factors) and tolerating a noise rate η=2ε/(l+2ε)−Δ. We complement this result by proving that this sample size is also necessary for any class C of VC-dimension d.

We then discuss the performance of the minimum disagreement strategy under both malicious and random classification noise models. For malicious noise we show an algorithm that, using deterministic hypotheses, learns unions of d intervals on the continuous domain [0, 1) using a sample size significantly smaller than that needed by the minimum disagreement strategy. For classification noise we show, generalizing a result by Laird, that order of d/εΔ ²) training examples suffice (up to logarithmic factors) to learn by minimizing disagreements any target class of VC-dimension d tolerating random classification noise rate η=1/2−Δ. Using a lower bound by Simon, we also prove that this sample size bound cannot be significantly improved.