Approximation Polynomial-Time Algorithms for Minimum Cost Intervention in Causal Effect Identification

Thumbnail

Event details

Date 29.08.2023
Hour 14:0016:00
Speaker Sepehr Elahi
Location
Category Conferences - Seminars
EDIC candidacy exam
Exam president: Prof. Michael Gastpar
Thesis advisor: Prof. Patrick Thiran
Thesis co-advisor: Prof. Negar Kiyavash
Co-examiner: Prof. Ola Svensson

Abstract
We investigate two fundamental problems in causality: causal discovery and causal identifiability. For the former, we summarize the important findings of two works, the first of which focuses on learning a causal graph through small-sized interventions, and the second of which looks at the problem of learning a causal graph through single-target noisy interventions. For the latter, we discuss a work that formulated the problem of ensuring identifiability through minimum cost interventions as a combinatorial optimization problem. We also briefly present our own work on the problem of minimum cost intervention for causal effect identification, where we derive bounds on the performance of a polynomial-time algorithm on random graphs.

Background papers
  1. Learning Causal Graphs with Small Interventions (2015)
  2. Sample Efficient Active Learning of Causal Trees (2019)
  3. Minimum Cost Intervention Design for Causal Effect Identification (2022)

Practical information

  • General public
  • Free

Tags

EDIC candidacy exam

Share