Communication Efficient Distributed Learning for Kernelized Contextual Bandits

Authors: Chuanhao Li, Huazheng Wang, Mengdi Wang, Hongning Wang

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We performed extensive empirical evaluations on both synthetic and real-world datasets, and the results (averaged over 3 runs) are reported in Figure 1, 2 and 3, respectively.
Researcher Affiliation Academia 1University of Virginia 2Oregon State University 3Princeton University {cl5ev,hw5x}@virginia.edu huazheng.wang@oregonstate.edu mengdiw@princeton.edu
Pseudocode Yes Algorithm 1 Distributed Kernel UCB (Dis Kernel UCB); Algorithm 2 Approximated Distributed Kernel UCB (Approx-Dis Kernel UCB); Algorithm 3 Ridge Leverage Score Sampling (RLS)
Open Source Code No Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [No]
Open Datasets Yes We performed extensive empirical evaluations on both synthetic and real-world datasets... Figure 2: Experiment results on UCI datasets. [8] Dheeru Dua and Casey Graff. UCI machine learning repository, 2017. Figure 3: Experiment results on Movie Lens & Yelp datasets. [13] F Maxwell Harper and Joseph A Konstan. The movielens datasets: History and context.
Dataset Splits No The paper mentions using grid search for hyperparameters ('grid search for α in {0.1, 1, 4}'), which implies some form of validation, but it does not explicitly provide details about specific training, validation, or test dataset splits in the main text.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running the experiments. It states '[N/A]' for the question regarding compute resources in the self-assessment.
Software Dependencies No The paper mentions using a 'Gaussian kernel' but does not specify any software libraries or frameworks (e.g., TensorFlow, PyTorch, scikit-learn) with their version numbers that were used for implementation.
Experiment Setup Yes For all the kernelized algorithms, we used the Gaussian kernel k(x, y) = exp( γ x y 2). We did a grid search of γ {0.1, 1, 4} for kernelized algorithms, and set D = 20 for Dis Lin UCB and Dis Kernel UCB, D = 5 for Approx-Dis Kernel UCB. For all algorithms, instead of using their theoretically derived exploration coefficient α, we followed the convention [20, 32] to use grid search for α in {0.1, 1, 4}.