Ancestral Causal Inference

Authors: Sara Magliacane, Tom Claassen, Joris M. Mooij

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We prove soundness and asymptotic consistency of our method and demonstrate that it can outperform the state-of-the-art on synthetic data, achieving a speedup of several orders of magnitude. We illustrate its practical feasibility by applying it to a challenging protein data set.
Researcher Affiliation Academia Sara Magliacane VU Amsterdam & University of Amsterdam sara.magliacane@gmail.com Tom Claassen Radboud University Nijmegen tomc@cs.ru.nl Joris M. Mooij University of Amsterdam j.m.mooij@uva.nl
Pseudocode No The paper mentions that ASP encoding is provided in the Supplementary Material but does not contain pseudocode or an algorithm block in the main text.
Open Source Code Yes Finally, we provide an open-source version of our algorithms and the evaluation framework, which can be easily extended, at http://github.com/caus-am/aci.
Open Datasets Yes Application on real data We consider the challenging task of reconstructing a signalling network from flow cytometry data [21] under different experimental conditions. [...] [21] K. Sachs, O. Perez, D. Pe er, D. Lauffenburger, and G. Nolan. Causal protein-signaling networks derived from multiparameter single-cell data. Science, 308:523 529, 2005.
Dataset Splits No The paper describes generating synthetic data and using a bootstrapping procedure for other methods, but it does not provide specific training/validation/test splits, percentages, or cross-validation details for its own model (ACI).
Hardware Specification Yes In Figure 1(a) we show the average execution times on a single core of a 2.80GHz CPU
Software Dependencies Yes For ACI we use the state-of-the-art ASP solver clingo 4 [6].
Experiment Setup Yes For the frequentist weights we use tests based on partial correlations and Fisher s z-transform to obtain approximate p-values (see, e.g., [9]) with significance level α = 0.05. [...] For the Bayesian weights, we use the Bayesian test for conditional independence presented in [13] as implemented by HEJ with a prior probability of 0.1 for independence. [...] We perform the bootstrap by repeating the following procedure 100 times: sample randomly half of the data, perform the independence tests, run Anytime (C)FCI. [...] For 8 variables HEJ can complete only four of the first 40 simulated models before the timeout of 2500s.