Multi-Task Learning for Contextual Bandits

Authors: Aniket Anand Deshmukh, Urun Dogan, Clay Scott

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test our algorithm on synthetic data and some multi-class classification datasets. We conclude in Section 6.
Researcher Affiliation Collaboration Aniket Anand Deshmukh Department of EECS University of Michigan Ann Arbor Ann Arbor, MI 48105 aniketde@umich.edu Urun Dogan Microsoft Research Cambridge CB1 2FB, UK urun.dogan@skype.net
Pseudocode Yes Algorithm 1 KMTL-UCB
Open Source Code No The paper does not provide concrete access to source code for the methodology described.
Open Datasets Yes We test our algorithm on synthetic data and some multi-class classification datasets. In the case of multi-class datasets, the number of arms N is the number of classes and the reward is 1 if we predict the correct class, otherwise it is 0. We consider the following datasets: Digits (N = 10, d = 64), Letter (N = 26, d = 16), MNIST (N = 10, d = 780 ), Pendigits (N = 10, d = 16), Segment (N = 7, d = 19) and USPS (N = 10, d = 256).
Dataset Splits Yes We separate the data into two parts validation set and test set. We use all Gaussian kernels and pre-select the bandwidth of kernels using five fold cross-validation on a holdout validation set and we use β = 0.1 for all experiments.
Hardware Specification No The paper does not provide specific hardware details used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details with version numbers.
Experiment Setup Yes We use all Gaussian kernels and pre-select the bandwidth of kernels using five fold cross-validation on a holdout validation set and we use β = 0.1 for all experiments.