Estimating Heterogeneous Treatment Effects: Mutual Information Bounds and Learning Algorithms
Authors: Xingzhuo Guo, Yuchen Zhang, Jianmin Wang, Mingsheng Long
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we propose detailed experiments to show the efficiency and generality of our methods. Code and data are available at https://github.com/thuml/Mit Net. We conduct simulation studies to identify the regularization effect of mutual information (MI). We performed numerical experiments on three real-world semi-synthetic datasets with binary or multinoulli treatments in order to illustrate the efficacy and generality of Mit Net. |
| Researcher Affiliation | Academia | 1School of Software, Tsinghua University. 2Institute for Interdisciplinary Information Sciences, Tsinghua University. Xingzhuo Guo <gxz19@mails.tsinghua.edu.cn>. Correspondence to: Mingsheng Long <mingsheng@tsinghua.edu.cn>. |
| Pseudocode | Yes | Algorithm 1 Mit Net with general treatments. |
| Open Source Code | Yes | Code and data are available at https://github.com/thuml/Mit Net. |
| Open Datasets | Yes | IHDP The Infant Health and Development Program (IHDP) dataset (Hill, 2011); News The News dataset was first proposed as a benchmark for counterfactual inference by Johansson et al. (2016) and was extended to the multinoulli treatment setting by Schwab et al. (2019); TCGA The Cancer Genomic Atlas (TCGA) project collected gene expression data from various types of cancers in 9659 individuals with 20531 covariates (Weinstein et al., 2013). IHDP and News are under MIT License. TCGA is licensed under a Creative Commons Attribution-Non Commercial-Share Alike 3.0 Unported License. |
| Dataset Splits | Yes | We randomly split the generated dataset to train, validation and test parts according to the ratio of 63/27/10. All experiments are performed with 63/27/10 train/validation/test splits. |
| Hardware Specification | No | The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers like Python 3.8, CPLEX 12.4) needed to replicate the experiment. |
| Experiment Setup | No | The paper mentions aspects of the experimental setup such as choosing 'optimal α and learning rate η' and using a 'three-layer neural network' with 'early stopping', but does not provide the concrete numerical values for these hyperparameters or detailed system-level training settings. |