Fast Proxy Experiment Design for Causal Effect Identification
Poster presentation at NeurIPS 2024
Presented by Sepehr Elahi & Sina Akbari at NeurIPS 2024.
| Poster time | Poster location | Paper PDF | Poster | Video |
|---|---|---|---|---|
| Wed 11 Dec 11 a.m. — 2 p.m. | West Ballroom A-D (#5000) | Link | Link | Link |
TL;DR: We present novel, highly efficient algorithms for designing minimum-cost proxy experiments (interventions) to identify causal effects, significantly outperforming the s.o.t.a.
The running time of our algorithm compared with the previous s.o.t.a. from Akbari et al., 2022 on random Erdos-Renyi acyclic directed mixed graphs with varying (bi)/directed sparsities for different number of variables (left); and different (bi)/directed edge probabilities (right).