Deep Multi-Modal Structural Equations For Causal Effect Estimation With Unstructured Proxies

Authors: Shachi Deshpande, Kaiwen Wang, Dhruv Sreenivas, Zheng Li, Volodymyr Kuleshov

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

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically demonstrate that our approach outperforms existing methods based on propensity scores and corrects for confounding using unstructured inputs on tasks in genomics and healthcare. Our methods can potentially support the use of large amounts of data that were previously not used in causal inference. 5 Experimental Results
Researcher Affiliation Academia Shachi Deshpande1,2, Kaiwen Wang1,2, Dhruv Sreenivas2, Zheng Li1,2, Volodymyr Kuleshov1,2 Department of Computer Science, Cornell Tech1 and Cornell University2 {ssd86, kw437, ds844, zl634, kuleshov}@cornell.edu
Pseudocode No The paper does not include pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes IHDP The Infant Health and Development Project (IHDP) is a popular benchmark for causal inference algorithms [13] that contains the outcomes of comprehensive early interventions for premature, low birth weight infants. We used the 1000 Human Genomes [3] dataset to generate a simulated multi-modal GWAS dataset. real-world plant GWAS dataset from the 1001 Genomes Project for Arabidopsis Thaliana plants [19, 55]
Dataset Splits No The paper mentions using "80% of the data points" for fitting models and discusses "training and test sets" (Section 5.1) and "Train+Val" and "Test" splits (Table 2). However, it does not provide specific percentages or counts for distinct train, validation, and test splits within the main text, instead deferring details to appendices.
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments, such as GPU or CPU models.
Software Dependencies No The paper does not specify software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup No The paper does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs) in the main text, instead referring to appendices for details.