A Hybrid Neural Network Model with Non-linear Factorization Machines for Collaborative Recommendation
In recent years, deep learning models have proven able to learn effective representation in many applications. However, the exploration of deep learning on recommender systems are relatively little. Although some recent work has utilized deep learning models to make recommendation, they primarily em...
Saved in:
Published in | Information Retrieval pp. 213 - 224 |
---|---|
Main Authors | , , , |
Format | Book Chapter |
Language | English |
Published |
Cham
Springer International Publishing
|
Series | Lecture Notes in Computer Science |
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | In recent years, deep learning models have proven able to learn effective representation in many applications. However, the exploration of deep learning on recommender systems are relatively little. Although some recent work has utilized deep learning models to make recommendation, they primarily employed it to learn abstract representation of auxiliary information and used matrix factorization to model the interactions between user and item features. Especially, the application of deep learning models to learn user-item interaction function is very new and there are few attempts to this direction. In this paper, we propose a novel model Non-Linear Factorization Machine (NLFM) for modelling user-item interaction function and a hybrid deep model named AE-NLFM for collaborative recommendation. NLFM leverages neural networks to learn non-linear feature interaction and is more expressive than FM [15]. Extensive experiments on three real-world datasets show that our proposed AE-NLFM significantly outperforms the state-of-the-art methods. |
---|---|
ISBN: | 3030010112 9783030010119 |
ISSN: | 0302-9743 1611-3349 |
DOI: | 10.1007/978-3-030-01012-6_17 |