Batched Nonparametric Bandits via k-Nearest Neighbor UCB
Authors: Sakshi Arya
TMLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical evaluations on synthetic and realworld datasets demonstrate that Ba Nk-UCB consistently outperforms binning-based baselines. Empirical results on synthetic and real data show consistent improvements over binning-based methods. |
| Researcher Affiliation | Academia | Sakshi Arya EMAIL Department of Mathematics, Applied Mathematics, and Statistics Case Western Reserve University |
| Pseudocode | Yes | Algorithm 1 Ba Nk-UCB for Batched Nonparametric Bandits |
| Open Source Code | No | The paper does not provide concrete access to source code. There is no mention of a code repository link, explicit code release statement, or code provided in supplementary materials for the methodology described. |
| Open Datasets | Yes | We evaluate the performance of Ba Nk-UCB and Ba SEDB algorithm on three publicly available classification datasets: (a) Rice (Cammeo & Osmancik, 2020), consisting of 3810 samples with 7 morphological features used to classify two rice varieties; (b) Occupancy Detection (Candanedo & Feldheim, 2016), with 8143 samples and 5 environmental sensor features used to predict room occupancy; and (c) EEG Eye State (Biermann, 2014), with 14980 samples and 14 EEG measurements used to classify eye state. |
| Dataset Splits | No | The paper mentions evaluating on real datasets using "a windowed average over 30 independent random permutations of each dataset" but does not specify explicit training, validation, or test splits (e.g., percentages or sample counts) typically needed for reproduction. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types, or memory amounts) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiment. |
| Experiment Setup | Yes | We set T = 10000, L = 1 for the Lipschitz constant in Assumption 3. We fix the number of batches to M = 5 to balance between frequent updates and computational efficiency, but the results remain consistent across different choices of M. ... For our proposed Ba Nk-UCB algorithm, we choose the same batch grid for a fair comparison. |