Nested Bandits
Authors: Matthieu Martin, Panayotis Mertikopoulos, Thibaud Rahier, Houssam Zenati
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We verify that this is indeed the case in a range of synthetic experiments in Section 5. In this section we present a series of numerical experiments designed to test the efficiency of our method compared to EXP3. |
| Researcher Affiliation | Collaboration | Matthieu Martin 1 Panayotis Mertikopoulos 2 1 Thibaud Rahier 1 Houssam Zenati 1 3 All authors in alphabetical order. 1Criteo AI Lab 2Univ. Grenoble Alpes, CNRS, Inria, Grenoble INP, LIG, 38000 Grenoble, France 3Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK, 38000 Grenoble, France. |
| Pseudocode | Yes | Algorithm 1: Nested exponential weights (NEW) |
| Open Source Code | Yes | The code to reproduce the experiments can be found at https://github.com/criteo-research/Nested-Exponential-Weights. |
| Open Datasets | No | We use a synthetic environment where we simulate nested similarity partitions with trees. The paper describes how data is generated in this synthetic environment but does not refer to a publicly available or open dataset with concrete access information like a link, DOI, or formal citation. |
| Dataset Splits | No | The paper mentions 'synthetic experiments' and discusses generating data for simulation, but it does not provide specific details about train/validation/test dataset splits, percentages, or methodologies for partitioning data for reproducibility. |
| Hardware Specification | Yes | All experiments were run on a Mac book pro laptop, with 1 processor of 6 cores @2.6GHz (6-Core Intel Core i7). |
| Software Dependencies | No | The paper discusses the implementation of algorithms and provides a GitHub link, but it does not specify any software dependencies with version numbers (e.g., Python, PyTorch, specific libraries). |
| Experiment Setup | Yes | Indeed, a decaying rate of 1/t was used for the score updates for all methods, as is common in the bandit litterature [18]. |