Adaptive Online Experimental Design for Causal Discovery

Authors: Muhammad Qasim Elahi, Lai Wei, Murat Kocaoglu, Mahsa Ghasemi

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct a series of experiments using random DAGs and the SACHS Bayesian network from bnlibrary (Scutari, 2009) to compare our algorithm with other baselines. The results show that our algorithm outperforms the baselines, requiring fewer samples.
Researcher Affiliation Academia 1School of Electrical and Computer Engineering, Purdue University, West Lafayette, Indiana, USA 2Life Sciences Institute, University of Michigan, Ann Arbor, Michigan, USA.
Pseudocode Yes Algorithm 1: Track-and-stop Causal Discovery
Open Source Code Yes The code to reproduce our experimental results and for running the baseline algorithms and our track-and-stop discovery algorithm is available at https://github.com/CausalML-Lab/Track-and-Stop-Discovery.
Open Datasets Yes We also evaluate the performance of causal discovery algorithms using the SACHS Bayesian network from the Discrete Bayesian Networks Repository in the bnlearn library (Scutari, 2009).
Dataset Splits No The paper describes generating graphs and evaluating performance but does not specify explicit training, validation, or test dataset splits (e.g., percentages or sample counts).
Hardware Specification No The paper mentions running 'simulations' and 'experiments' but does not specify any particular hardware components like GPU or CPU models, or cloud computing resources used.
Software Dependencies No The paper mentions using the 'Causal Discovery Toolbox (Kalainathan et al., 2020)' and 'bnlearn library (Scutari, 2009)' but does not provide specific version numbers for these software dependencies.
Experiment Setup No The paper describes the data generation process and the overall algorithm but does not specify concrete hyperparameters (e.g., learning rate, batch size, epochs) or detailed training configurations for the algorithm itself.