Instrumental Variable Regression with Confounder Balancing

Authors: Anpeng Wu, Kun Kuang, Bo Li, Fei Wu

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that our algorithm outperforms the existing approaches.
Researcher Affiliation Collaboration 1Department of Computer Science and Technology, Zhejiang University, China 2School of Economics and Managemen, Tsinghua University, China 3Shanghai Institute for Advanced Study, Zhejiang University, China 4Shanghai AI Laboratory, China.
Pseudocode Yes The details of pseudo-code (Algorithm 1) and the network structures (Table 5) of CB-IV are provided in Section E.1 in Appendix.
Open Source Code Yes 2The code is available at: https://github.com/anpwu/CB-IV
Open Datasets Yes IHDP3: The Infant Health and Development Program (IHDP) [...] 3http://www.fredjo.com/ [...] Twins4: Twins dataset is derived from all twins born in the USA between the years 1989 and 1991 (Almond et al., 2005). [...] 4http://www.nber.org/data/
Dataset Splits Yes We conduct our experiments over the 100 realizations of IHDP and 10 realizations of Twins with a 63/27/10 proportion of train/validation/test splits.
Hardware Specification Yes Hardware used: Ubuntu 16.04.5 LTS operating system with 2 * Intel Xeon E5-2678 v3 CPU, 384GB of RAM, and 4 * Ge Force GTX 1080Ti GPU with 44GB of VRAM.
Software Dependencies Yes Software used: Python with Tensor Flow 1.15.0, Num Py 1.17.4, and Matplot Lib 3.1.1.
Experiment Setup Yes Table 5 shows the details of the structure networks of CB-IV in different datasets. [...] Learning Rate 0.0005 Optimizer Adam α 0.01/0.001