CAUSE: Learning Granger Causality from Event Sequences using Attribution Methods
Authors: Wei Zhang, Thomas Panum, Somesh Jha, Prasad Chalasani, David Page
ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we present the experiments that are designed to evaluate CAUSE and answer the following three questions: Goodness-of-Fit: How good is CAUSE at fitting multitype event sequences? Causality Discovery: How accurate is CAUSE at discovering Granger causality between event types? Scalability: How scalable is CAUSE? |
| Researcher Affiliation | Collaboration | 1Computer Scineces Department, University of Wisconsin Madison, Madison, WI, USA. 2Department of Electronic Systems, Aalborg University, Aalborg, Denmark. 3Xai Pient Inc., Princeton, NJ, USA. 4Department of Biostatistics and Bioinformatics, Duke University, Durham, NC, USA. |
| Pseudocode | Yes | Algorithm 1: Computation of the Granger causality statistic. |
| Open Source Code | Yes | Reproducibility. We publish our data and our code at https://github.com/razhangwei/CAUSE. |
| Open Datasets | Yes | IPTV (Luo et al., 2015): Each sequence records the history of TV watching behavior of a user, and the event types are the TV program categories. This dataset, however, does not contain ground-truth causality between TV program categories. Meme Tracker (MT):3 Each sequence represents how a phrase or quote appeared on various online websites over time during the period of August 2008 to April 2009, and the event types are the domains of the top websites. Like previous studies (Achab et al., 2018; Xiao et al., 2019), a weighted ground-truth causality matrix was approximated by whether one site contains any URLs linking to another site. |
| Dataset Splits | Yes | We performed five-fold cross-validation and report the average results. |
| Hardware Specification | No | The paper mentions 'modern computation hardware (such as GPUs)' and 'the GPU we tested on' in Section 3.3, and 'GPU we tested on' in Section 4.2.3, but it does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for experiments. |
| Software Dependencies | No | The paper does not provide specific software dependency details with version numbers. It refers to 'Tick: a Python library' in its references but does not specify the version of Python or any other libraries used for implementation or experiments. |
| Experiment Setup | No | The paper states: 'The implementation details and hyperparameter configurations for CAUSE and various baselines are provided in Appendix C.2'. These details are not present in the main body of the paper. |