Decentralized Gradient-Quantization Based Matrix Factorization for Fast Privacy-Preserving Point-of-Interest Recommendation

Authors: Xuebin Zhou, Zhibin Hu, Jin Huang, Jian Chen

JAIR 2023 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results conducted on real-world POI datasets demonstrate the effectiveness of the proposed algorithm. We conduct experiments on two real-world POI datasets, Gowalla and Foursquare (Cho, Myers, & Leskovec, 2011; Yang, Zhang, & Qu, 2016).
Researcher Affiliation Academia Xuebin Zhou EMAIL South China University of Technology Guangzhou, Guangdong, 510000, China; Zhibin Hu EMAIL South China Normal University Guangzhou, Guangdong, 510000, China; Jin Huang EMAIL (Corresponding Author) South China Normal University Guangzhou, Guangdong, 510000, China; Jian Chen EMAIL (Corresponding Author) South China University of Technology Guangzhou, Guangdong, 510000, China
Pseudocode Yes Algorithm 1 DGMF Optimization Input: training data (D), learning rate (η), regularization strength (α, β, γ, δ, µ), the number of neighbors (N), and maximum iterations (T). Output: user latent matrix (W), shared item latent tensor (P), and specific item latent tensor (Q). 1: For u = 1 to U do 2: Initialize wu, P u, Qu.
Open Source Code No The paper does not provide concrete access to source code for the methodology described. There are no explicit statements about code release, repository links, or mentions of code in supplementary materials.
Open Datasets Yes We conduct experiments on two real-world POI datasets, Gowalla and Foursquare (Cho, Myers, & Leskovec, 2011; Yang, Zhang, & Qu, 2016).
Dataset Splits Yes Table 1 shows the statistics after processing two datasets, with which we randomly sample 90% as training set and the rest 10% as test set, and Figure 2 illustrates their geological aggregation patterns.
Hardware Specification No The paper does not provide specific hardware details (such as GPU or CPU models, memory, or cluster specifications) used for running the experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library names with version numbers, that would be needed to replicate the experiment.
Experiment Setup Yes For all experiments, we set the length of ranked POI list k {5, 10}, and search the learning rate η in [10 3, 10 1]. User regularizer α and the item regularizers β, γ, δ, µ are determined from [10 5, 10 1]. For the latent vector dimension K, we vary its values in {5, 10, 15}. ... We vary the maximum number of neighbors N in {0, 1, 10, 50, 100}. Based on the training set, we use 5fold cross validation to choose the hyper-parameters.