Factorization Bandits for Interactive Recommendation

Authors: Huazheng Wang, Qingyun Wu, Hongning Wang

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experimentations on both simulations and large-scale real-world datasets confirmed the advantages of the proposed algorithm compared with several state-of-the-art factorization-based and bandit-based collaborative filtering methods.
Researcher Affiliation Academia Huazheng Wang, Qingyun Wu, Hongning Wang Department of Computer Science University of Virginia, Charlottesville VA, 22904 USA {hw7ww,qw2ky,hw5x}@virginia.edu
Pseudocode Yes Algorithm 1 Factor UCB
Open Source Code No The paper does not provide an unambiguous statement of releasing source code or a direct link to a code repository.
Open Datasets Yes Yahoo dataset: This data set contains 10-days clickstream logs from Yahoo! Today Module collected in May 2009, totalling 45,811,883 user visits (Li et al. 2010). Last FM dataset: This dataset is extracted from the online music streaming service Last.fm (http://www.last.fm). It contains 1,892 users, 17,632 items (artists), and the users social network graph.
Dataset Splits No The paper describes splitting users into 'learning group' and 'testing group' but does not explicitly mention a 'validation' set or specific train/validation/test splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory, or cloud instance types) used for running experiments.
Software Dependencies No The paper does not provide specific software names with version numbers for replication.
Experiment Setup Yes We fixed the dimension d of observable features to 20, the dimension l of latent item factors to 5, user size N to 100, the standard derivation σ of Gaussian noise to 0.1, and the item pool size K to 1000 in our simulation. ... in our empirical evaluations the algorithm s performance is not sensitive to this setting. ... As a result, in all our following experiments we will use the manually set αu t and αa t.