On the Parameter Identifiability of Partially Observed Linear Causal Models

Authors: Xinshuai Dong, Ignavier Ng, Biwei Huang, Yuewen Sun, Songyao Jin, Roberto Legaspi, Peter Spirtes, Kun Zhang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical studies on both synthetic and real-world datasets validate our identifiability theory and the effectiveness of the proposed method in the finitesample regime.
Researcher Affiliation Collaboration Xinshuai Dong1* Ignavier Ng1* Biwei Huang2 Yuewen Sun3 Songyao Jin3 Roberto Legaspi4 Peter Spirtes1 Kun Zhang1,3 1Carnegie Mellon University 2University of California San Diego 3Mohamed bin Zayed University of Artificial Intelligence 4KDDI Research
Pseudocode No The paper describes methods and objectives for parameter estimation but does not include structured pseudocode or algorithm blocks.
Open Source Code Yes Code: https://github.com/dongxinshuai/scm-identify.
Open Datasets Yes In this section, we employ a famous psychometric dataset Big Five dataset https: //openpsychometrics.org/, to validate our method.
Dataset Splits No The paper mentions using different sample sizes (2k, 5k, 10k) for synthetic data and a dataset of 20,000 data points for real-world data, but it does not explicitly provide training/validation/test dataset splits needed to reproduce the experiments.
Hardware Specification Yes We conduct all the experiments with single Intel(R) Xeon(R) CPU E5-2470.
Software Dependencies No The paper states 'Our code is based on Python3.7 and Py Torch [37]' which provides a version for Python but not for PyTorch, and does not list multiple key software components with their specific versions.
Experiment Setup Yes Data is standardized and the optimization in Eqs. (4), (5), and (7) are solved by Adam [27], with a learning rate of 0.02. We will rely on 30 random starts and choose the one with the best likelihood.