Applications of Common Entropy for Causal Inference
Authors: Murat Kocaoglu, Sanjay Shakkottai, Alexandros G. Dimakis, Constantine Caramanis, Sriram Vishwanath
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our algorithms on synthetic and real data to establish their performance. |
| Researcher Affiliation | Collaboration | Murat Kocaoglu MIT-IBM Watson AI Lab, IBM Research, Sanjay Shakkottai The University of Texas at Austin, Alexandros G. Dimakis The University of Texas at Austin, Constantine Caramanis The University of Texas at Austin, Sriram Vishwanath The University of Texas at Austin |
| Pseudocode | Yes | Algorithm 1 Latent Search: Iterative Update Algorithm, Algorithm 2 Infer Graph: Identifying the Latent Graph, Algorithm 3 Entropic PC (for F = False) and Entropic PC-C (for F = True) |
| Open Source Code | No | The paper does not include an unambiguous statement or a direct link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | On ADULT Dataset: We compare the performance of Entropic PC with the baseline PC algorithm that we used for our modifications. ... [13] Dua Dheeru and EfiKarra Taniskidou. UCI machine learning repository, 2017. ... Second, in Figure 3a, we run Latent Search on the real cause-effect pairs from Tuebingen dataset [36]. ... [36] Joris M Mooij, Jonas Peters, Dominik Janzing, Jakob Zscheischler, and Bernhard Schölkopf. Distinguishing cause from effect using observational data: methods and benchmarks. The Journal of Machine Learning Research, 17(1):1103 1204, 2016. |
| Dataset Splits | No | The paper mentions generating "datasets with 1k, 5k, 80k samples" for synthetic data, and uses real datasets like ADULT and Tuebingen, but does not explicitly provide specific training/validation/test splits (e.g., percentages or exact counts) or refer to predefined standard splits for all datasets used. |
| Hardware Specification | No | The paper mentions using "computing resources from TACC" in the acknowledgements, but it does not provide specific details on the hardware used, such as GPU/CPU models, memory, or other detailed computer specifications for running the experiments. |
| Software Dependencies | No | The paper states: "We use pcalg package in R [24, 18]". However, it does not specify the version numbers for the `pcalg` package or R, which is required for reproducibility. |
| Experiment Setup | Yes | When using Latent Search to approximate Rényi1 common entropy, we will run it for multiple β values and pick the distribution q(.) with the smallest H(Z) such that I(X; Y |Z) θ for a practical threshold θ to declare conditional independence. ... To address this, in simulations H( ) is set to 0.1 min{H(X), H(Y )} in line 10. This and the choice of 0.8 as the coefficient in line 16 can be seen as hyper-parameters to be tuned. ... we conclude that 0.001 is a suitable CMI threshold for this dataset. |