Reducing Balancing Error for Causal Inference via Optimal Transport

Authors: Yuguang Yan, Hao Zhou, Zeqin Yang, Weilin Chen, Ruichu Cai, Zhifeng Hao

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experiments on both synthetic and real-world datasets demonstrate the effectiveness of our proposed method. ... We conduct experiments on both synthetic and real-world datasets to evaluate the performance of our proposed method. ... We compare the performance of OICL with the following methods: ... For evaluating the performance of the conducted methods, we adopt the mean absolute errors (MAE) as the metric. We carry out the experiments 10 times and report the mean and standard deviation. The results are reported in Table 1.
Researcher Affiliation Academia 1School of Computer Science, Guangdong University of Technology, Guangzhou, China 2Pazhou Laboratory (Huangpu), Guangzhou, China 3College of Science, Shantou University, Shantou, China. Correspondence to: Ruichu Cai <cairuichu@gmail.com>.
Pseudocode Yes Algorithm 1 Optimal transport for causal Inference by Cost Learning (OICL).
Open Source Code Yes Our code is available at https://github.com/ygyan/OICL.
Open Datasets Yes La Londe consists of two parts. The first part comes from the RCT (NSW), and we replace the control group in NSW with another control group from the observational data (PSID) in the second part. ... https://users.nber.org/ rdehejia/data/.nswdata2.html. Twins is collected from the twins born in USA between 1989-1991 (Almond et al., 2005). ... Infant Health and Development Program (IHDP) aims to study the treatment effect of specialist home visits on infants future cognitive test scores. ... We consider setting A in the NPCI package (Dorie, 2016).
Dataset Splits No The paper describes generating simulation data and using real-world datasets but does not explicitly provide training, validation, or test dataset splits (e.g., percentages, sample counts, or citations to predefined splits) in the main text.
Hardware Specification Yes Hardware used in this experiment are: CPU: Intel i5-12600K, GPU: NVIDIA Ge Force RTX 4090.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment.
Experiment Setup Yes We take La Londe as an example to evaluate the effects of the parameters of our model. We vary the parameters λc, λt, γc, γt in Eqs. (29), from 10^-3 to 5 and plot the result in Figure 1.