Abstract 

Population risk is always of primary interest in machine learning; however, learning algorithms only have access to the empirical risk. Even for applications with nonconvex nonsmooth losses (such as modern deep networks), the population risk is generally significantly more well-behaved from an optimization point of view than the empirical risk. In particular, sampling can create many spurious local minima. We consider a general framework which aims to optimize a smooth nonconvex function F (population risk) given only access to an approximation f (empirical risk) that is pointwise close to F  Our objective is to find the -approximate local minima of the underlying function F while avoiding the shallow local minima—arising because of the tolerance —which exist only in f . We propose a simple algorithm based on stochastic gradient descent (SGD) on a smoothed version of f that is guaranteed to achieve our goal as long as We also provide an almost matching lower bound showing that our algorithm achieves optimal error tolerance among all algorithms making a polynomial number of queries of f . As a concrete example, we show that our results can be directly used to give sample complexities for learning a ReLU unit.