Data-faithful Feature Attribution: Mitigating Unobservable Confounders via Instrumental Variables
Authors: Qiheng Sun, Haocheng Xia, Jinfei Liu
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experiments on both synthetic and real-world datasets demonstrate the effectiveness of our approaches. |
| Researcher Affiliation | Academia | Qiheng Sun1,2, Haocheng Xia3, Jinfei Liu1,2 1Zhejiang University 2Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security 3Siebel School of Computing and Data Science University of Illinois Urbana-Champaign {qiheng_sun,jinfeiliu}@zju.edu.cn, hxia7@illinois.edu |
| Pseudocode | Yes | For the algorithm process we propose, which utilizes confidence intervals to optimize the calculation of Shapley values, refer to Algorithm 1. |
| Open Source Code | Yes | Our code can be found in the repository at https://github.com/ZJU-DIVER/IV-SHAP. |
| Open Datasets | Yes | The first real dataset we used is the Griliches76 dataset [17, 36]... The second real dataset we use is the Angrist and Krueger dataset [3]... |
| Dataset Splits | No | The paper describes generating synthetic data and training models, but it does not provide specific train/validation/test dataset splits (e.g., percentages, sample counts, or citations to predefined splits) to reproduce the partitioning. |
| Hardware Specification | Yes | We conduct experiments on a machine with 2 Montage(R) Jintide(R) C6226R @ 2.90GHz and 256GB memory. |
| Software Dependencies | No | The paper mentions software components like 'neural network model' and 'XGBoost', but it does not specify any version numbers for these or other libraries/packages. |
| Experiment Setup | No | The paper mentions using neural networks and XGBoost models, and discusses loss functions, but it does not provide specific hyperparameters such as learning rate, batch size, number of epochs, or detailed optimizer settings for model training. |