Targeted Reduction of Causal Models

Armin Kekić, Bernhard Schölkopf, Michel Besserve
11/30/2023
stat.MLcs.LG

Official code: akekic/targeted-causal-reduction (1 star, MIT, last commit 2024-06-14).

Abstract

Why does a phenomenon occur? Addressing this question is central to most scientific inquiries and often relies on simulations of scientific models. As models become more intricate, deciphering the causes behind phenomena in high-dimensional spaces of interconnected variables becomes increasingly challenging. Causal Representation Learning (CRL) offers a promising avenue to uncover interpretable causal patterns within these simulations through an interventional lens. However, developing general CRL frameworks suitable for practical applications remains an open challenge. We introduce Targeted Causal Reduction (TCR), a method for condensing complex intervenable models into a concise set of causal factors that explain a specific target phenomenon. We propose an information theoretic objective to learn TCR from interventional data of simulations, establish identifiability for continuous variables under shift interventions and present a practical algorithm for learning TCRs. Its ability to generate interpretable high-level explanations from complex models is demonstrated on toy and mechanical systems, illustrating its potential to assist scientists in the study of complex phenomena in a broad range of disciplines.

View on arXivView PDF

Code Implementations(1)

akekic/targeted-causal-reductionOfficial
MIT

TCR is a method for explaining a phenomenon (called target) in high-dimensional simulations (low-level model) by learning a low-dimensional causal model (high-level model) that captures the most important causes of the target. It uses shift interventions in the low-level model and its effects on the target to learn the high-level model.

13PythonMay 28, 20242 years agoMIT

Similar Papers

On the Identifiability of Causal Abstractions

89% match2025

Identifiable Causal Representation Learning: Unsupervised, Multi-View, and Multi-Environment

87% match2024

Implicit Causal Representation Learning via Switchable Mechanisms

85% match2024

Unifying Causal Representation Learning with the Invariance Principle

85% match2024

Nonparametric Identifiability of Causal Representations from Unknown Interventions

85% match1 repo262026

Causal Representation Learning Made Identifiable by Grouping of Observational Variables

84% match1 repo72026

Cite this paper

@article{keki2023targeted,
  title  = {Targeted Reduction of Causal Models},
  author = {Armin Kekić and Bernhard Schölkopf and Michel Besserve},
  year   = {2023},
  eprint = {2311.18639},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2311.18639v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2311.18639.svg)](https://codeofpaper.com/paper/2311.18639)

Discussion