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. 2012.13976
9
9

Intervention Efficient Algorithms for Approximate Learning of Causal Graphs

27 December 2020
Raghavendra Addanki
A. Mcgregor
Cameron Musco
    CML
ArXivPDFHTML
Abstract

We study the problem of learning the causal relationships between a set of observed variables in the presence of latents, while minimizing the cost of interventions on the observed variables. We assume access to an undirected graph GGG on the observed variables whose edges represent either all direct causal relationships or, less restrictively, a superset of causal relationships (identified, e.g., via conditional independence tests or a domain expert). Our goal is to recover the directions of all causal or ancestral relations in GGG, via a minimum cost set of interventions. It is known that constructing an exact minimum cost intervention set for an arbitrary graph GGG is NP-hard. We further argue that, conditioned on the hardness of approximate graph coloring, no polynomial time algorithm can achieve an approximation factor better than Θ(log⁡n)\Theta(\log n)Θ(logn), where nnn is the number of observed variables in GGG. To overcome this limitation, we introduce a bi-criteria approximation goal that lets us recover the directions of all but ϵn2\epsilon n^2ϵn2 edges in GGG, for some specified error parameter ϵ>0\epsilon > 0ϵ>0. Under this relaxed goal, we give polynomial time algorithms that achieve intervention cost within a small constant factor of the optimal. Our algorithms combine work on efficient intervention design and the design of low-cost separating set systems, with ideas from the literature on graph property testing.

View on arXiv
Comments on this paper