Create and explore imaginative 3D worlds with our new product Dreamfield (opens in a new tab)
Papers
Topics
Authors
Recent
Search
2000 character limit reached

Learning Powers of Poisson Binomial Distributions

Published 18 Jul 2017 in cs.DS, cs.LG, math.ST, and stat.TH | (1707.05662v1)

Abstract: We introduce the problem of simultaneously learning all powers of a Poisson Binomial Distribution (PBD). A PBD of order nn is the distribution of a sum of nn mutually independent Bernoulli random variables XiX_i, where E[Xi]=pi\mathbb{E}[X_i] = p_i. The kk'th power of this distribution, for kk in a range [m][m], is the distribution of Pk=∑i=1nXi(k)P_k = \sum_{i=1}^n X_i^{(k)}, where each Bernoulli random variable Xi(k)X_i^{(k)} has E[Xi(k)]=(pi)k\mathbb{E}[X_i^{(k)}] = (p_i)^k. The learning algorithm can query any power PkP_k several times and succeeds in learning all powers in the range, if with probability at least 1−δ1- \delta: given any k∈[m]k \in [m], it returns a probability distribution QkQ_k with total variation distance from PkP_k at most ϵ\epsilon. We provide almost matching lower and upper bounds on query complexity for this problem. We first show a lower bound on the query complexity on PBD powers instances with many distinct parameters pip_i which are separated, and we almost match this lower bound by examining the query complexity of simultaneously learning all the powers of a special class of PBD's resembling the PBD's of our lower bound. We study the fundamental setting of a Binomial distribution, and provide an optimal algorithm which uses O(1/ϵ2)O(1/\epsilon^2) samples. Diakonikolas, Kane and Stewart [COLT'16] showed a lower bound of Ω(21/ϵ)\Omega(2^{1/\epsilon}) samples to learn the pip_i's within error ϵ\epsilon. The question whether sampling from powers of PBDs can reduce this sampling complexity, has a negative answer since we show that the exponential number of samples is inevitable. Having sampling access to the powers of a PBD we then give a nearly optimal algorithm that learns its pip_i's. To prove our two last lower bounds we extend the classical minimax risk definition from statistics to estimating functions of sequences of distributions.

Summary

No one has generated a summary of this paper yet.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.