Online Corrupted User Detection and Regret Minimization

Authors: Zhiyong Wang, Jize Xie, Tong Yu, Shuai Li, John C.S. Lui

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental With extensive experiments, our methods achieve superior performance over previous bandit algorithms and high corrupted user detection accuracy.
Researcher Affiliation Collaboration Zhiyong Wang The Chinese University of Hong Kong zywang21@cse.cuhk.edu.hk Jize Xie Shanghai Jiao Tong University xjzzjl@sjtu.edu.cn Tong Yu Adobe Research worktongyu@gmail.com Shuai Li Shanghai Jiao Tong University shuaili8@sjtu.edu.cn John C.S. Lui The Chinese University of Hong Kong cslui@cse.cuhk.edu.hk
Pseudocode Yes Algorithm 1 RCLUB-WCU
Open Source Code No The paper does not provide explicit statements or links for the open-source code of the described methodology.
Open Datasets Yes We use three real-world data Movielens [11], Amazon[31], and Yelp [33].
Dataset Splits No The paper describes data generation and processing, but does not explicitly mention train/validation/test dataset splits with percentages or counts.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory specifications).
Software Dependencies No The paper does not specify software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes We use u = 1, 000 users and m = 10 clusters, where each cluster contains 100 users. We randomly select 100 users as the corrupted users. The preference and arm (item) vectors are drawn in d 1 (d = 50) dimensions with each entry a standard Gaussian variable and then normalized, added one more dimension with constant 1, and divided by 2 [21]. We fix an arm set with |A| = 1000 items, at each round, 20 items are randomly selected to form a set At to choose from. Following [40, 3], in the first k rounds, we always flip the reward of corrupted users by setting rt = x T atθit,t + ηt. And we leave the remaining T k rounds intact. Here we set T = 1, 000, 000 and k = 20, 000.