Deep Counterfactual Estimation with Categorical Background Variables

Authors: Edward De Brouwer

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate our construction on three different datasets with different high-dimensional modalities (images and time series) and demonstrate accurate counterfactual estimation.
Researcher Affiliation Academia Edward De Brouwer ESAT-STADIUS KU Leuven edward.debrouwer@esat.kuleuven.be
Pseudocode Yes We also present a pseudo-code description of the procedure in Algorithm 1.
Open Source Code Yes Our code is available at https://github.com/edebrouwer/cfqp.
Open Datasets No As counterfactual inference evaluation on real-world data is a complex and ongoing research area, we use synthetic datasets inspired by case studies from the literature. Details about the data generation are given in Appendix A.
Dataset Splits Yes In Figure 4a, we show the reconstruction error on the validation set (i.e., on the factual data) and the counterfactual reconstruction error of CFQP for different values of K on the image dataset with true number of latent classes K0 = 6.
Hardware Specification No The paper states that 'The total amount of compute and the type of resources used is available in the Appendix Section B,' but the specific Appendix B content is not provided in the paper excerpt, preventing confirmation of explicit hardware details like GPU/CPU models.
Software Dependencies No The paper indicates that 'The licenses of all software packages used in our experiments is available at https://anonymous.4open.science/r/cfqp/licenses.txt,' but it does not explicitly list software packages with their specific version numbers within the main text or the provided excerpt.
Experiment Setup No The paper outlines the training procedure (Initialization, Expectation, Maximization) and mentions 'epochs' in Algorithm 1, and the checklist indicates that 'training details in Section B of the Appendix,' but specific hyperparameter values (e.g., learning rate, batch size) are not explicitly detailed in the provided main text.