A Generative Adversarial Framework for Bounding Confounded Causal Effects

Authors: Yaowei Hu, Yongkai Wu, Lu Zhang, Xintao Wu12104-12112

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments using both synthetic and real-world datasets are conducted.
Researcher Affiliation Academia 1 University of Arkansas 2 Clemson University yaoweihu@uark.edu, yongkaw@clemson.edu, lz006@uark.edu, xintaowu@uark.edu
Pseudocode Yes The pseudocode of above procedure is given in Algorithm 1 in the supplementary file.
Open Source Code Yes Reproducibility. The source code is available at https://github.com/yaoweihu/Bound-Confounded-Causal-Effects.
Open Datasets Yes We use both synthetic data and a real-world dataset, Adult (Dheeru and Karra Taniskidou 2017).
Dataset Splits No The paper does not provide specific training, validation, or test set splits (e.g., percentages or counts) for the datasets used in the experiments.
Hardware Specification No The paper does not specify any hardware details such as GPU models, CPU types, or memory used for running the experiments.
Software Dependencies No The paper mentions software components like "Wasserstein GAN" and "adaptive gradient clipping" but does not provide specific version numbers for any libraries, frameworks, or programming languages.
Experiment Setup Yes In the implementation of our method, we use one hidden layer with 16 nodes for all generators GV ( ) and neural networks h V ( ; ), and use Re LU as the activation function. [...] The threshold η in the constraint is set to 0.001, and we take 50 solutions satisfying the constraint to compute the mean and variance. The upper bound is computed as mean + std, and the lower bound is computed as mean std. The training procedure is as follows. We continually sample mini-batches of noise samples z.