Causal Discovery from Subsampled Time Series with Proxy Variables

Authors: Mingzhou Liu, Xinwei Sun, Lingjing Hu, Yizhou Wang

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we evaluate our method on synthetic data and a real-world application, i.e., discovering causal pathways in Alzheimer s disease. ... We use the F1-score, precision, and recall, where precision and recall (resp.) measure the accuracy and completeness of identified causal edges, and F1 = 2 precision recall / precision+recall.
Researcher Affiliation Academia 1 School of Computer Science, Peking University 2 Center on Frontiers of Computing Studies (CFCS), Peking University 3 School of Data Science, Fudan University 4 Yanjing Medical College, Capital Medical University 5 Institute for Artificial Intelligence, Peking University
Pseudocode Yes Algorithm 1: Discover the summary graph
Open Source Code Yes Our code is available at https://github.com/lmz123321/proxy_causal_discovery.
Open Datasets Yes We consider the Alzheimer s Disease Neuroimaging Initiative (ADNI) dataset [4], in which the imaging data is acquired from structural Magnetic Resonance Imaging (s MRI) scans. [4] Ronald Carl Petersen, Paul S Aisen, Laurel A Beckett, Michael C Donohue, Anthony Collins Gamst, Danielle J Harvey, Clifford R Jack, William J Jagust, Leslie M Shaw, Arthur W Toga, et al. Alzheimer s disease neuroimaging initiative (adni): clinical characterization. Neurology, 74(3):201 209, 2010.
Dataset Splits No The paper describes sample sizes for synthetic data and number of subjects for ADNI, but does not provide specific train/validation/test splits (e.g., percentages or exact counts for each split).
Hardware Specification No The paper does not provide specific details on the hardware used for running experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper mentions using the 'causallearn package' for MAG recovery but does not specify its version number or any other software dependencies with version information.
Experiment Setup Yes The significance level is set to 0.05. For each graph, we generate temporal data with the structural equation Xi(t) = j PAi fij(Xj(t 1)) + Ni, where the function fij is randomly chosen from {linear,sin,tanh,sqrt}, the exogenous noise Ni is randomly sampled from {uniform,gauss.,exp.,gamma}.