Noisy population recovery in polynomial time
Abstract
In the noisy population recovery problem of Dvir et al., the goal is to learn an unknown distribution f on binary strings of length n from noisy samples. For some parameter μ ∈ [0,1], a noisy sample is generated by flipping each coordinate of a sample from f independently with probability (1-μ)/2. We assume an upper bound k on the size of the support of the distribution, and the goal is to estimate the probability of any string to within some given error . It is known that the algorithmic complexity and sample complexity of this problem are polynomially related to each other. We show that for μ > 0, the sample complexity (and hence the algorithmic complexity) is bounded by a polynomial in k, n and 1/ improving upon the previous best result of poly(k k,n,1/) due to Lovett and Zhang. Our proof combines ideas from Lovett and Zhang with a noise attenuated version of M\"obius inversion. In turn, the latter crucially uses the construction of robust local inverse due to Moitra and Saks.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.