Non-Stationary Bandits with Auto-Regressive Temporal Dependency
Authors: Qinyi Chen, Negin Golrezaei, Djallel Bouneffouf
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we conduct a real-world case study on tourism demand prediction [36] in Section 7, confirming the superiority of AR2 compared to benchmark algorithms. There, we also show that the techniques and high-level ideas of our algorithm can be readily extended to handle more complicated, rapidly changing temporal structure (e.g., general AR-p processes) while still achieving good performance. Our case study is complemented by synthetic experiments in Appendix A which show the strength of AR2 against a number of benchmarks designed for stationary and non-stationary settings. |
| Researcher Affiliation | Collaboration | Qinyi Chen Operations Research Center Massachusetts Institute of Technology qinyic@mit.edu Negin Golrezaei Sloan School of Management Massachusetts Institute of Technology golrezae@mit.edu Djallel Bouneffouf IBM Research djallel.bouneffouf@ibm.com |
| Pseudocode | Yes | Algorithm 1 Alternating and Restarting algorithm for non-stationary AR bandits (AR2) |
| Open Source Code | No | The paper does not contain an explicit statement or a link indicating that the source code for the described methodology (AR2) is publicly available. |
| Open Datasets | Yes | The international tourism demand dataset [36], obtained from the Australian Bureau of Statistics, records the number of individual tourist arrivals to Australia from Hong Kong during each quarter between the years 1975-1989. |
| Dataset Splits | No | The paper describes running experiments over a total number of rounds (e.g., T=10,000 or T=200) and mentions using 'Test rounds' for AR parameter estimation, but it does not specify explicit training, validation, or test dataset splits, percentages, or methodology for data partitioning for model evaluation. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory, or cloud instance types) used to run the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers (e.g., Python version, specific libraries like PyTorch or TensorFlow with their versions) that would be needed to replicate the experiments. |
| Experiment Setup | Yes | Algorithm 1 (AR2) and Algorithm 2 (AR2-p) explicitly list input parameters such as 'AR Parameter α, stochastic rate of change σ, epoch size ep, parameter c0' and their derivations. Section 7 and Appendix A describe specific settings for k, T, σ, and how αi and σi are drawn for numerical studies. |