Ask a Question

Prefer a chat interface with context about you and your work?

Minimax Rates of Entropy Estimation on Large Alphabets via Best Polynomial Approximation

Minimax Rates of Entropy Estimation on Large Alphabets via Best Polynomial Approximation

Consider the problem of estimating the Shannon entropy of a distribution over k elements from n independent samples. We show that the minimax mean-square error is within the universal multiplicative constant factors of (k/n log k) <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> t log <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> k/n if n exceeds a constant …