Chebyshev polynomials, moment matching, and optimal estimation of the unseen

Abstract

We consider the problem of estimating the support size of a discrete distribution whose minimum nonzero mass is at least1k . Under the independent sampling model, we show that the sample complexity, that is, the minimal sample size to achieve an additive error of εk with probabilityatleast 0.1 is within universal constant factors oflogkk log2 1ε , which improves the state-of-the-art result ofε2 logkk in [In Advances in Neural Information Processing Systems (2013) 2157–2165]. Similar characterization of the minimax risk is also obtained. Our procedure is a linear estimator based on the Chebyshev polynomial and its approximation-theoretic properties, which can be evaluated in O(n + log2 k) time and attains the sample complexity within constant factors. The superiority of the proposed estimator in terms of accuracy, computational efficiency and scalability is demonstrated in a variety of synthetic and real datasets.