The Adaptive Doubly Robust Estimator and a Paradox Concerning Logging Policy
Authors: Masahiro Kato, Kenichiro McAlinn, Shota Yasui
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We confirm this hypothesis through simulation studies. |
| Researcher Affiliation | Collaboration | Masahiro Kato Cyber Agent, Inc. Kenichiro Mc Alinn Temple University Shota Yasui Cyber Agent, Inc. |
| Pseudocode | No | The paper describes algorithms and methods mathematically and in prose but does not include any formal pseudocode blocks or algorithm listings. |
| Open Source Code | Yes | We have the instructions to reproduce the result in the appendix and readme file of the supplementary material (we included source code and requirements). |
| Open Datasets | Yes | From the LIBSVM repository (Chang & Lin, 2011), we use the mnist, satimage, sensorless, and connect-4 datasets. |
| Dataset Splits | No | The paper mentions sample sizes used for generating synthetic datasets and for benchmark datasets (e.g., 'sample sizes 800, 1,000, and 1,200'), but it does not specify explicit training, validation, or test splits (e.g., percentages or counts for each subset) for any of the datasets used in experiments. |
| Hardware Specification | No | The paper explicitly states in its ethics checklist: 'The computations are very simple and could be done on a personal laptop, so resource information is not included.' |
| Software Dependencies | No | The paper mentions using 'kernelized Ridge least squares and logistic regression' for estimation, and specific MAB algorithms (Lin UCB, Lin TS), but it does not provide specific version numbers for any software libraries, programming languages, or tools used to implement these methods. |
| Experiment Setup | Yes | For estimating f and the logging policy, we use the kernelized Ridge least squares and logistic regression, respectively. We use a Gaussian kernel, and both hyper-parameters of the Ridge and kernel are chosen from {0.01, 0.1, 1}. We conduct ten experiments by changing sample sizes and MAB algorithms. For the sample size T(3), we use 100, 250, 500, 750, and 10, 000 with the Lin UCB and Lin TS algorithms. For the sample sizes T(1) and T(2), we use 1, 000 and 100, 000 respectively. |