SIGMA: Selective Gated Mamba for Sequential Recommendation

Authors: Ziwei Liu, Qidong Liu, Yejing Wang, Wanyu Wang, Pengyue Jia, Maolin Wang, Zitao Liu, Yi Chang, Xiangyu Zhao

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrate that SIGMA significantly outperforms existing baselines across five real-world datasets.
Researcher Affiliation Academia 1City University of Hong Kong 2School of Auto. Science & Engineering, MOEKLINNS Lab, Xi an Jiaotong University 3Jinan University 4Jilin University
Pseudocode No The paper describes the methodology using mathematical formulas and descriptive text in sections like 'Methodology' and 'G-Mamba Block', but no explicitly labeled pseudocode or algorithm blocks are present.
Open Source Code Yes Code https://github.com/Applied-Machine-Learning Lab/SIMGA
Open Datasets Yes We conduct comprehensive experiments on five representative real-world datasets i.e., Yelp1, Amazon series2 (Beauty, Sports and Games) and Movie Lens-1M3.The statistics of datasets after preprocessing are shown in Table 1. ... 1https://www.yelp.com/dataset 2https://cseweb.ucsd.edu/jmcauley/datasets.html#amazon reviews 3https://grouplens.org/datasets/movielens/
Dataset Splits No The paper describes categorizing users into subsets based on interaction length for analysis ('Short (0 5), Medium (5 20), and Long (20+)'), but it does not explicitly provide details on how the datasets are split into training, validation, and test sets for model evaluation, such as specific percentages or sample counts.
Hardware Specification Yes For GPU selection, all experiments are conducted on a single NVIDIA L4 GPU.
Software Dependencies No The paper mentions using the Adam optimizer and setting a learning rate, but it does not specify versions for any programming languages, libraries, or other software components used in the implementation.
Experiment Setup Yes The Adam optimizer (Kingma and Ba 2014) is used with a learning rate of 0.001. For a fair comparison, the embedding dimension for all tested models is set to 64.