New metrics and search algorithms for weighted causal DAGs

Authors: Davin Choo, Kirankumar Shiragur

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

Reproducibility Variable Result LLM Response
Research Type Experimental Some empirical results are shown in Section 5 and source code is provided in the supplementary materials. and We implement and benchmark ALG-GENERALIZED. (Section 5)
Researcher Affiliation Academia 1School of Computing, National University of Singapore 2Broad Institute of MIT and Harvard.
Pseudocode Yes Algorithm 1 Atomic weighted adaptive search., Algorithm 2 Resolve Dangling, Algorithm 3 ALG-GENERALIZED, Algorithm 4 Resolve Dangling Generalized, Algorithm 5 Clique Intervention, Algorithm 6 Naive weighted adaptive search.
Open Source Code Yes source code is provided in the supplementary materials. (Section 1) and Source code implementation and experimental scripts are available at https://github.com/cxjdavin/new-metrics-and-search-algorithms-for-weighted-causal-DAGs. (Appendix G. Experiments).
Open Datasets No The paper uses synthetic graph classes (Erdos-Renyi styled graphs and Tree-like graphs) which are generated for the experiments. It describes the parameters and generation process, but does not provide access to pre-generated datasets or state that these synthetic datasets are publicly available. It also does not use any commonly known public datasets.
Dataset Splits No The paper describes the generation of synthetic graphs for experiments but does not provide specific details on how these graphs are split into training, validation, or test sets, or specify cross-validation methods.
Hardware Specification Yes All our experiments are conducted on an Ubuntu server with two AMD EPYC 7532 CPU and 256GB DDR4 RAM.
Software Dependencies No The paper mentions that experiments were conducted on an 'Ubuntu server' but does not specify any software dependencies like programming languages, libraries, or frameworks with their respective version numbers.
Experiment Setup Yes We ran experiments for α {0, 1} and β = 1 on two different types of weight classes for a graph on n vertices... Type 1. Vertex weights are i.i.d. sampled from an exponential distribution exp(n2)... Type 2. A random p = 0.1 fraction of vertices are assigned weight n2... Experiment 1 Graph class 1 with n {10, 15, 20, 25} and density ρ = 0.1.