Sub-Interest-Aware Representation Uniformity for Recommender System

Authors: Ruijia Ma, Yahong Lian, Chunyao Song

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on four datasets demonstrate that SIURec achieves superior learning of uniformity (with an average improvement of 4.26% in accuracy compared to eleven SOTA methods) and exhibits robustness across different hyperparameter settings. [...] 5 Experiments In this section, we evaluate SIURec on different datasets to answer following questions: RQ1: How does SIURec perform compared to other competitive methods under different experimental settings? RQ2: How does each component of SIURec contribute to performance enhancement? RQ3: How robust is SIURec under various parameter settings?
Researcher Affiliation Academia Ruijia Ma, Yahong Lian, Chunyao Song* College of Computer Science, TJ Key Lab of NDST, DISSec, TMCC, TBI Center, Nankai University, Tianjin, China EMAIL, EMAIL
Pseudocode No The paper describes the methodology using mathematical equations and textual explanations, but it does not include a clearly labeled pseudocode or algorithm block.
Open Source Code Yes Code https://github.com/xderui/SIURec
Open Datasets Yes Datasets. We select four commonly used public benchmark datasets in our experiments: Movie Lens-1M (ML1M), Gowalla, Amazon-Beauty (Beauty) and Amazon-Book (Book). The dataset statistics are shown in Table 1.
Dataset Splits Yes For each dataset, we group them by user and divide them into 8:1:1 ratios for training, validation, and testing.
Hardware Specification Yes All experiments are implemented on an Intel(R) Xeon(R) Silver 4110 @ 2.10GHz CPU and an NVIDIA Ge Force RTX 2080 Ti GPU.
Software Dependencies No The paper mentions Light GCN as a base model and other comparative methods, but it does not specify the versions of the programming languages or libraries used for the implementation of SIURec.
Experiment Setup Yes Regard to SIURec, we set the initial values αB = 1, αU = 1, and αR = 2.5 10 5. For each baseline, we set the parameters following the suggestions from each individual s work.