GRACE-C: Generalized Rate Agnostic Causal Estimation via Constraints

Authors: Mohammadsajad Abavisani, David Danks, Sergey Plis

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

Reproducibility Variable Result LLM Response
Research Type Experimental We generated 50 random single-SCC graphs each of 8, 16, and 32 nodes, all with average degree of 1.4 outgoing edges per node. We then undersampled each graph by 2, 3, and 4, and used each individual undersampled graph as input to s RASL (i.e., 150 different input graphs for each size).
Researcher Affiliation Academia Mohammadsajad Abavisani Department of Electrical and Computer Engineering Georgia Institute of Technology Atlanta, GA 30332 s.abavisani@gatech.edu David Danks Halicioglu Data Science Institute and Department of Philosophy University of California San Diego San Diego, CA 92093 ddanks@ucsd.edu Sergey M. Plis TRe NDS center Department of Computer Science Georgia State University Atlanta, GA 30302 s.m.plis@gmail.com
Pseudocode Yes Listing 1: s RASL encoding in the clingo ASP-language
Open Source Code Yes The full code is available at https://gitlab.com/undersampling/gunfolds
Open Datasets Yes we used publicly available data from (Sanchez-Romero et al., 2019) and applied our method on the resting-state f MRI data.
Dataset Splits No The paper mentions using "10 datasets of concatenated recording for 10 individuals, comprising seven regions of interest from medial temporal lobe, each containing 4,210 datapoints" but does not specify any training, validation, or test dataset splits.
Hardware Specification Yes These experiments used Clingo in parallel mode using 10 threads computing on AMD EPYC 7551 CPUs. Given computational complexity, all experiments were run on a slurm cluster that submits jobs to one of the 19 machines on the same network, each with 64 cores and 512 GB of RAM.
Software Dependencies No The paper mentions 'Clingo' and 'ASP-language' but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes We used a 24-hour timeout (i.e., stopped the run if it did not finish in 24 hours).