Invariant Ancestry Search
Authors: Phillip B Mogensen, Nikolaj Thams, Jonas Peters
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We develop scalable algorithms and perform experiments on simulated and real data. We evaluate our method in several simulation studies as well as a real-world data set on gene perturbations. |
| Researcher Affiliation | Academia | 1Department of Mathematical Sciences, University of Copenhagen, Denmark. |
| Pseudocode | Yes | Algorithm 1 An algorithm for computing SIAS from data |
| Open Source Code | Yes | Code is provided at https://github.com/PhillipMogensen/InvariantAncestrySearch. |
| Open Datasets | Yes | We evaluate our approach in a data set on gene expression in yeast (Kemmeren et al., 2014). |
| Dataset Splits | No | The paper describes generating synthetic datasets and uses a real-world dataset, but it does not provide specific details on train/validation/test splits (percentages, counts, or methodology) for reproducing its experiments. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory, or cloud instance types) used for running its experiments. |
| Software Dependencies | No | We provide a function for listing all minimally invariant sets in our python code; it uses an implementation of the above mentioned algorithm, provided in the R (R Core Team, 2021) package dagitty (Textor et al., 2016). Unknown-Target Interventional Greedy Sparsest Permutation (UT-IGSP) (Squires et al., 2020) using the implemention from the Python package Causal DAG. (These mention software packages but do not provide specific version numbers for them.) |
| Experiment Setup | Yes | Throughout the section, we consider a significance level of α = 5%. For a detailed description of the simulations, see Appendix E.2. When d = 6, we test hypotheses with a correction factor C = 3 6/3 = 9...When d = 100, we test hypotheses with the correction factor C(1) of Theorem 5.5. In both cases, we test the hypothesis of invariance of the empty set at level α0 = 10 6 |