Optimal Kernel Choice for Score Function-based Causal Discovery

Authors: Wenjie Wang, Biwei Huang, Feng Liu, Xinge You, Tongliang Liu, Kun Zhang, Mingming Gong

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments on both synthetic data and realworld benchmarks, and the results demonstrate that our proposed method outperforms heuristic kernel selection methods.
Researcher Affiliation Academia 1School of Mathematics and Statistics, The University of Melbourne, Australia 2Department of Machine Learning, Mohamed bin Zayed University of Artificial Intelligence, United Arab Emirates 3Halicio glu Data Science Institute (HDSI), University of California, San Diego, United States 4School of Computing and Information Systems, The University of Melbourne, Australia 5Huazhong University of Science and Technology, China 6School of Computer Science, Faculty of Engineering, The University of Sydney, Australia 7Department of Philosophy, Carnegie Mellon University, United States.
Pseudocode No No explicit pseudocode or algorithm blocks labeled 'Algorithm' or 'Pseudocode' were found.
Open Source Code No The paper states 'Our code is implemented in Python 3.8.7 and PyTorch 1.11.0.' and refers to implementations of baseline methods (e.g., 'causal-learn package', 'NOTEARS2', 'NS-MLP2', 'DAG-GNN3', 'Dibs4') but does not provide a direct link or explicit statement about the public availability of their *own* source code for the proposed method.
Open Datasets Yes We further evaluated our method on two widely-used causal discovery benchmarks: SACH and CHILD networks. The SACH network comprises 11 variables and 17 edges, while the CHILD network consists of 20 variables with 25 edges. ... We randomly selected data with sample sizes of n = 200,500,1000 and 2000, repeating 20 times for each sample size.
Dataset Splits No The paper describes using cross-validation for a baseline method (CV) with '10-fold cross validation' but does not provide explicit training, validation, or test dataset splits, nor does it describe cross-validation for its own proposed method.
Hardware Specification No The paper states, 'This research was undertaken using the LIEF HPC-GPGPU Facility hosted at the University of Melbourne. This Facility was established with the assistance of LIEF Grant LE170100200.' However, it does not specify any particular GPU models, CPU models, memory, or other detailed hardware specifications.
Software Dependencies Yes Our code is implemented in Python 3.8.7 and Py Torch 1.11.0.
Experiment Setup Yes To avoid numerical issues, we explicitly set the ranges for these parameters as σx,σp [0.1,10] and σε [0.001,10]. We employed the L-BFGS (Liu & Nocedal, 1989) as the optimization method for our model with the default hyper-parameter setting.