EE-Net: Exploitation-Exploration Neural Networks in Contextual Bandits

Authors: Yikun Ban, Yuchen Yan, Arindam Banerjee, Jingrui He

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on four real-world datasets, showing that EE-Net outperforms existing linear and neural contextual bandit baselines on real-world datasets.
Researcher Affiliation Academia Yikun Ban, Yuchen Yan, Arindam Banerjee, Jingrui He University of Illinois at Urbana-Champaign {yikunb2, yucheny5, arindamb, jingrui}@illinois.edu
Pseudocode Yes Algorithm 1 EE-Net
Open Source Code Yes Codes are available at 1. 1https://github.com/banyikun/EE-Net-ICLR-2022
Open Datasets Yes We use four real-world datasets: Mnist, Yelp, Movielens, and Disin, the details and settings of which are attached in Appendix A. ... MNIST dataset. MNIST is a well-known image dataset (Le Cun et al., 1998)... Yelp and Movielens (Harper and Konstan, 2015) datasets. ... Disin (Ahmed et al., 2018) dataset.
Dataset Splits No The paper describes an online learning bandit problem where data is processed sequentially over T rounds. It does not provide explicit training, validation, and test splits with percentages or counts, as is typical for static supervised learning datasets.
Hardware Specification No The paper does not specify the hardware used for experiments, such as particular CPU or GPU models, or cloud computing instances with their specifications.
Software Dependencies No The paper does not specify software dependencies with version numbers (e.g., Python, PyTorch versions, or other libraries).
Experiment Setup Yes For all neural networks, we conduct the grid search for learning rate over (0.01, 0.001, 0.0005, 0.0001). For all grid-searched parameters, we choose the best of them for the comparison and report the averaged results of 10 runs for all methods. ... For all the neural-based methods including EE-Net, the exploitation network f1 is built by a 2-layer fully-connected network with 100 width. For the exploration network f2, we use a 2-layer fully-connected network with 100 width as well.