Causal Inference via Kernel Deviance Measures

Authors: Jovana Mitrovic, Dino Sejdinovic, Yee Whye Teh

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test our method on real-world time series data and the real-world benchmark dataset Tübingen Cause-Effect Pairs where we outperform state-of-the-art approaches. 4 Experimental Results
Researcher Affiliation Collaboration Jovana Mitrovic Dino Sejdinovic Yee Whye Teh Department of Statistics, University of Oxford [mitrovic, dino.sejdinovic, y.w.teh]@stats.ox.ac.uk Now at Deep Mind, UK.
Pseudocode No Algorithms summarizing our causal inference methodology are given in the supplementary material.
Open Source Code No The paper mentions that 'Algorithms summarizing our causal inference methodology are given in the supplementary material', but does not explicitly state that the code is open-source or provide a link for its own implementation.
Open Datasets Yes For this purpose, we test KCDC on the only widely used benchmark dataset Tübingen Cause-Effect Pairs (TCEP) [15]. This dataset is comprised of real-world cause-effect samples that are collected across very diverse subject areas with the true causal direction provided by human experts. J. M. Mooij, D. Janzing, J. Zscheischler, and B. Schölkopf. Cause-effect pairs repository. 2015. http://webdav.tuebingen.mpg.de/cause-effect/.
Dataset Splits No The paper describes using synthetic data and the TCEP dataset, but does not provide specific details on train/validation/test splits (percentages, counts, or explicit splitting methodology) for reproducibility.
Hardware Specification No The paper does not provide any specific hardware details such as GPU/CPU models, memory, or specific cloud instances used for running experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., library names with specific versions, or solver versions) needed to replicate the experiments.
Experiment Setup Yes We report the results when using the log kernel on the input and the rational quadratic kernel on the response. both on the response and input we used either a log kernel k(x, x ) = log( x x q + 1) with q in [2, 3, 4] or an RBF kernel with bandwidth [1, 1.5, 2] times the median heuristic.