期刊
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
卷 33, 期 10, 页码 5775-5788出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TNNLS.2021.3071392
关键词
High-dimensional and sparse (HiDS) matrix; latent factor (LF) analysis; L-1 norm; L-2 norm; recommender system (RS)
类别
资金
- National Natural Science Foundation of China [61702475, 61772493, 62002337, 61902370]
- CAAI-Huawei MindSpore Open Fund [CAAIXSJLJJ2020-004B]
- Natural Science Foundation of Chongqing (China) [cstc2019jcyj-msxmX0578, cstc2019jcyjjqX0013]
- Chinese Academy of Sciences Light of West China Program
- Technology Innovation and Application Development Project of Chongqing, China [cstc2018jszx-cyzdX0041, cstc2019jscx-fxydX0027]
- Pioneer Hundred Talents Program of Chinese Academy of Sciences
This article introduces a novel LF model that combines the characteristics of L1 and L2 norms to improve accuracy and stability in handling HiDS data with outliers. Experimental results demonstrate that the model outperforms state-of-the-art models in predicting missing data, indicating its potential for real-world HiDS data applications.
A recommender system (RS) is highly efficient in filtering people's desired information from high-dimensional and sparse (HiDS) data. To date, a latent factor (LF)-based approach becomes highly popular when implementing a RS. However, current LF models mostly adopt single distance-oriented Loss like an L-2 norm-oriented one, which ignores target data's characteristics described by other metrics like an L-1 norm-oriented one. To investigate this issue, this article proposes an L-1-and-L-2-norm-oriented LF((LF)-F-3) model. It adopts twofold ideas: 1) aggregating L-1 norm's robustness and L-2 norm's stability to form its Loss and 2) adaptively adjusting weights of L-1 and L-2 norms in its Loss. By doing so, it achieves fine aggregation effects with L-1 norm-oriented Loss's robustness and L-2 norm-oriented Loss's stability to precisely describe HiDS data with outliers. Experimental results on nine HiDS datasets generated by real systems show that an (LF)-F-3 model significantly outperforms state-of-the-art models in prediction accuracy for missing data of an HiDS dataset. Its computational efficiency is also comparable with the most efficient LF models. Hence, it has good potential for addressing HiDS data from real applications.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据