Federated Causal Discovery from Heterogeneous Data

Authors: Loka Li, Ignavier Ng, Gongxu Luo, Biwei Huang, Guangyi Chen, Tongliang Liu, Bin Gu, Kun Zhang

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on synthetic and real datasets to show the efficacy of our method.
Researcher Affiliation Academia Loka Li1, Ignavier Ng2, Gongxu Luo1, Biwei Huang3, Guangyi Chen1,2, Tongliang Liu1, Bin Gu1, Kun Zhang1,2 1 Mohamed bin Zayed University of Artificial Intelligence 2 Carnegie Mellon University 3 University of California San Diego
Pseudocode Yes Algorithm 1 Fed CDH: Federated Causal Discovery from Heterogeneous Data
Open Source Code Yes The code is available at https://github.com/lokali/Fed CDH.git. Our source code has been appended in the Supplementary Materials.
Open Datasets Yes To evaluate our method, we conduct extensive experiments on synthetic datasets including linear Gaussian models and general functional models, and real-world dataset including f MRI Hippocampus (Poldrack et al., 2015) and HK Stock Market datasets (Huang et al., 2020).
Dataset Splits No The paper describes various experimental settings for evaluation (e.g., varying number of variables, clients, and samples), but does not explicitly mention a dedicated validation set or a train/validation/test split for model development or hyperparameter tuning.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, or memory specifications) used for running the experiments. It only generally refers to computational time.
Software Dependencies No The paper mentions software like 'Causal-learn package' and 'Python' but does not specify their version numbers or other crucial software dependencies with specific versions.
Experiment Setup Yes We evaluate variable d {6, 12, 18, 24, 30} while fixing other variables such as K=10 and nk=100. We set client K {2, 4, 8, 16, 32} while fixing others such as d=6 and nk=100. We let the sample size in one client nk {25, 50, 100, 200, 400} while fixing other variables such as d=6 and K=10. We set the hyperparameter h to 5, and set the significance level for FCIT to 0.05.