Robust Neural Contextual Bandit against Adversarial Corruptions

Authors: Yunzhe Qi, Yikun Ban, Arindam Banerjee, Jingrui He

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We also conduct experiments against baselines on real data sets under different scenarios, in order to demonstrate the effectiveness of our proposed R-Neural UCB.
Researcher Affiliation Academia University of Illinois at Urbana-Champaign Champaign, IL 61820
Pseudocode Yes Algorithm 1 Robust Neural-UCB (R-Neural UCB)
Open Source Code Yes We include the source code along with our submission.
Open Datasets Yes Movie Lens 20M rating data set [41], Amazon Recommendation data set [43], MNIST data set [56]
Dataset Splits No The paper uses well-known datasets but does not explicitly state the training, validation, or testing splits (e.g., percentages or counts) or reference standard splits from citations.
Hardware Specification Yes All experiments are conducted on a server with an Intel Xeon CPU and NVIDIA V100 GPUs.
Software Dependencies No The paper mentions using deep learning models but does not provide specific software dependencies with version numbers (e.g., PyTorch, TensorFlow versions).
Experiment Setup Yes For all UCB-based baselines, we choose the exploration parameter through grid search over the range {0.01, 0.1, 1}. We set L = 2 for all deep learning models, including our proposed Neural UCB-WGD and R-Neural UCB, and set the network width to m = 200. The learning rate for all neural algorithms is chosen by grid search from the range {0.0001, 0.001, 0.01}. For all methods, we select the regularization parameter λ from the range {0.0001, 0.001, 0.01}. The scaling parameter α for Neural UCB-WGD and R-Neural UCB is chosen from {0.2, 0.5, 1}.