A new learner achieves the statistically optimal risk bound for hypothesis classes with finite VC dimension d. The algorithm matches known lower bounds for agnostic PAC learning, providing a sample complexity that scales with the square root of the optimal risk and d/n.
HOW THIS AFFECTS YOU
●
researcherYou can use this construction to achieve optimal risk bounds in agnostic PAC learning settings.