Efficient Explorative Key-Term Selection Strategies for Conversational Contextual Bandits

Authors: Zhiyong Wang, Xutong Liu, Shuai Li, John C. S. Lui

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on synthetic and real-world data show significant advantages of our algorithms in learning accuracy (up to 54% improvement) and computational efficiency (up to 72% improvement), compared to the classic Con UCB algorithm, showing the potential benefit to recommender systems.
Researcher Affiliation Academia 1The Chinese University of Hong Kong ,Hong Kong SAR, China 2Shanghai Jiao Tong University, Shanghai, China
Pseudocode Yes Algorithm 1: General Con Lin UCB framework
Open Source Code Yes Codes are available at https://github.com/ZhiyongWangWzy/ConLinUCB.
Open Datasets Yes Last.FM is a dataset for artist recommendations containing 186,479 interaction records between 1,892 users and 17,632 artists. Movielens is a dataset for movie recommendation containing 47,957 interaction records between 2,113 users and 10,197 movies. ... The data is generated following (Li et al. 2019; Zhang et al. 2020; Wu et al. 2021).
Dataset Splits No The paper describes the total number of rounds (T) for the experiments and how the datasets were used, but it does not specify explicit training, validation, or test dataset splits.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as CPU or GPU models, or memory specifications.
Software Dependencies No The paper does not explicitly list specific software dependencies with their version numbers required for reproducibility.
Experiment Setup Yes Following (Zhang et al. 2020), we set T = 1,000, b(t) = 5 log(t+1) and |At| = 50, unless otherwise stated. ... We set b(t) = fq log t and vary fq to change the conversation frequencies, i.e., fq {5, 10, 20, 30}. We vary |At| to be 25, 50, 100, 200, 500.