Doubly Robust Proximal Causal Learning for Continuous Treatments
Authors: Yong Wu, Yanwei Fu, Shouyan Wang, Xinwei Sun
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the utility of our estimator on synthetic datasets and real-world applications. We demonstrate the utility and efficiency on synthetic data and the policy-making (Donohue III & Levitt, 2001). In this section, we evaluate the effectiveness of our method using two sets of synthetic data one in a low-dimensional context and the other in a high-dimensional context as well as the legalized abortion and crime dataset (Donohue III & Levitt, 2001). |
| Researcher Affiliation | Academia | 1Institute of Science and Technology for Brain-Inspired Intelligence, Fudan University 2School of Data Science, Fudan University 3Zhangjiang Fudan International Innovation Center 4Key Laboratory of Computational Neuroscience and Brain-Inspired Intelligence (Fudan University) 5MOE Frontiers Center for Brain Science, Fudan University 6Shanghai Engineering Research Center of AI & Robotics, Fudan University 7Engineering Research Center of AI & Robotics, Ministry of Education, Fudan University |
| Pseudocode | No | The paper describes algorithms and methods in text and mathematical formulations but does not include any explicit pseudocode blocks or algorithm listings. |
| Open Source Code | Yes | Code is available at https://github.com/yezichu/PCL_Continuous_Treatment. |
| Open Datasets | Yes | We obtain the data from Donohue III & Levitt (2001); Mastouri et al. (2021) that explores the relationship between legalized abortion and crime. The dataset is available at https://github.com/yuchen-zhu/kernel_proxies/tree/main/data/sim_1d_no_x. |
| Dataset Splits | No | The paper mentions 'Sample sizes' and reports 'c MSE across 100 equally spaced points' but does not provide specific train/validation/test dataset splits with percentages, counts, or explicit references to predefined splits. |
| Hardware Specification | No | The computations in this research were performed using the CFFF platform of Fudan University. This statement is too general and does not provide specific hardware details (e.g., CPU/GPU models, memory). |
| Software Dependencies | No | The paper mentions software components like 'Gaussian kernels', 'KDE', 'CNFs', 'Adam', and 'probaforms' but does not provide specific version numbers for any of them. |
| Experiment Setup | Yes | In the PKIPW and PKDR estimators, we choose the second-order Epanechnikov kernel, with bandwidth hbw = cˆσAn 1/5 with estimated std ˆσA and the hyperparameter c > 0. In our paper, we vary c over the range {0.5, 1, 1.5, , 4.0}. To estimate nuisance functions, we parameterize Q and M (resp., H and G) via RKHS for q0 (resp., h0), where we use Gaussian kernels with the bandwidth parameters being initialized using the median distance heuristic. For policy estimation, we employ the KDE in the low-dimensional synthetic dataset and the real-world data, while opting for CNFs in the highdimensional synthetic dataset. The section H.4 also details hyperparameters like learning rate, epochs, batch size, and weight decay for CNFs. |