Reproduce “Neural Exploitation and Exploration of Contextual Bandits” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yikun Ban, Yuchen Yan, Arindam Banerjee, Jingrui He
EE-Net learns signed prediction errors with a second neural network to adjust reward estimates in contextual bandits. It combines this learned correction with predicted rewards when choosing among options, allowing adjustments upward or downward. Across four dataset-derived benchmark tasks, it reports lower cumulative regret than comparison methods; on MNIST, regret is 842.3 versus NeuralUCB’s 943.5 after 10,000 rounds.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.