Learning Mixtures of Unknown Causal Interventions

Authors: Abhinav Kumar, Kirankumar Shiragur, Caroline Uhler

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct a simulation study to validate our theoretical findings. We show that as sample size increases, one can recover the mixture parameters, identify the unknown intervention targets, and learn the underlying causal graph with high accuracy.
Researcher Affiliation Collaboration Abhinav Kumar LIDS, Massachusetts Institute of Technology Broad Institute of MIT and Harvard akumar03@mit.edu Kirankumar Shiragur Microsoft Research kshiragur@microsoft.com Caroline Uhler LIDS, Massachusetts Institute of Technology Broad Institute of MIT and Harvard
Pseudocode Yes Algorithm 1: Mixture-UTIGSP
Open Source Code Yes The source code to all the experiments can be found in the following Git Hub repository: https://github.com/Big Bang0072/mixture_mec
Open Datasets Yes We evaluate our method on the Protein Signaling dataset [22] to demonstrate real-world applicability. ... For details see Wang et al. [31] and Sachs et al. [22].
Dataset Splits No The paper describes how the mixed dataset is generated and evaluated, but it does not specify explicit train/validation/test dataset splits with percentages or counts for its experiments.
Hardware Specification No The paper states: 'We use an internal cluster of CPUs to run all our experiments.' This does not provide specific hardware details such as CPU model, memory, or GPU specifications.
Software Dependencies No The paper states, 'we use the standard sklearn python package [19] that implements an EM algorithm to estimate the parameters of the mixture,' but it does not specify version numbers for Python or scikit-learn.
Experiment Setup Yes The initial noise distribution for all the nodes is univariate Gaussian distribution N(0, 1). ... we set it to a very small value of 10 9 for numerical stability. ... we use the default tol = 10 3 used by Gaussian Mixture... we use a cutoff threshold of 0.07... Specifically, we use α = 10 3 for both Memoized CITester and Memoized Invariance Tester functions used by UT-IGSP.