ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2002.05378
39
20
v1v2 (latest)

Efficient Distance Approximation for Structured High-Dimensional Distributions via Learning

13 February 2020
Arnab Bhattacharyya
Sutanu Gayen
Kuldeep S. Meel
N. V. Vinodchandran
ArXiv (abs)PDFHTML
Abstract

We design efficient distance approximation algorithms for several classes of structured high-dimensional distributions. Specifically, we show algorithms for the following problems: - Given sample access to two Bayesian networks P1P_1P1​ and P2P_2P2​ over known directed acyclic graphs G1G_1G1​ and G2G_2G2​ having nnn nodes and bounded in-degree, approximate dtv(P1,P2)d_{tv}(P_1,P_2)dtv​(P1​,P2​) to within additive error ϵ\epsilonϵ using poly(n,ϵ)poly(n,\epsilon)poly(n,ϵ) samples and time - Given sample access to two ferromagnetic Ising models P1P_1P1​ and P2P_2P2​ on nnn variables with bounded width, approximate dtv(P1,P2)d_{tv}(P_1, P_2)dtv​(P1​,P2​) to within additive error ϵ\epsilonϵ using poly(n,ϵ)poly(n,\epsilon)poly(n,ϵ) samples and time - Given sample access to two nnn-dimensional Gaussians P1P_1P1​ and P2P_2P2​, approximate dtv(P1,P2)d_{tv}(P_1, P_2)dtv​(P1​,P2​) to within additive error ϵ\epsilonϵ using poly(n,ϵ)poly(n,\epsilon)poly(n,ϵ) samples and time - Given access to observations from two causal models PPP and QQQ on nnn variables that are defined over known causal graphs, approximate dtv(Pa,Qa)d_{tv}(P_a, Q_a)dtv​(Pa​,Qa​) to within additive error ϵ\epsilonϵ using poly(n,ϵ)poly(n,\epsilon)poly(n,ϵ) samples, where PaP_aPa​ and QaQ_aQa​ are the interventional distributions obtained by the intervention do(A=a)do(A=a)do(A=a) on PPP and QQQ respectively for a particular variable AAA. Our results are the first efficient distance approximation algorithms for these well-studied problems. They are derived using a simple and general connection to distribution learning algorithms. The distance approximation algorithms imply new efficient algorithms for {\em tolerant} testing of closeness of the above-mentioned structured high-dimensional distributions.

View on arXiv
Comments on this paper