Double Cross & Deep Network for News Recommendation

DOI

10.2991/assehr.k.200401.026How to use a DOI?

Keywords

neural networks, feature crossing, deep learning, news recommendation

Abstract

News recommendation algorithms are widely used in many Internet products that people use. With the increase of commercial value, the research on various recommendation algorithms has become more and more interesting. This paper proposes the Double Cross & Deep Network (DCDN) algorithm, which is used in news recommendation. On the basis of the DCN network, the features of “relevant articles” involved in the field of news recommendation are separately extracted, and high-level intersections are made with user information and seed information, respectively. The parameters of the two Cross Network and Deep Network of the DCDN network are independent, and users can change the parameters according to the predicted demand. When the number of layers of the Cross Network is increased, the relevance of the recommendation can be increased, while the equivalent number of layers of the Deep Network can increase the diversity of recommendations. Experiments show that compared with DCN networks, DCDN networks have better parameter performance and faster model operation.

Copyright

© 2020, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Zhihong Yang
AU  - Yuewei Wu
AU  - Muqing Wu
AU  - Yulong Wang
PY  - 2020
DA  - 2020/04/06
TI  - Double Cross & Deep Network for News Recommendation
BT  - Proceedings of the International Conference on Education, Economics and Information Management (ICEEIM 2019)
PB  - Atlantis Press
SP  - 101
EP  - 106
SN  - 2352-5398
UR  - https://doi.org/10.2991/assehr.k.200401.026
DO  - 10.2991/assehr.k.200401.026
ID  - Yang2020
ER  -