本文的核心思想体现在文中图2。 首先我们有用户的行为数据和attributes，本文的做法是，利用encoder将用户行为压缩成attributes，同时再利用decoder将attributes重新构造成用户行为。其中，encoder和decoder之间的参数有一定的联系，如Symmetric 。最后，为了让强化一部分关系，以及弱化另一部分关系，作者在objective function引入了一个low-rank约束（很有意思，值得研究下）。
这样只要学出decoder，和new user的attribute，我们就可以reconstruct用户行为了。
当然，作者认为cold-start recommendation和图像领域的zero learning有着内在联系。比如他们解决的问题很像。所以，zero learning也是一种概念，并不是某种特定的技术。
| 文献题目 | 去谷歌学术搜索 | ||||||||||
| From Zero-Shot Learning to Cold-Start Recommendation | |||||||||||
| 文献作者 | Jingjing Li | ||||||||||
| 文献发表年限 | 2019 | ||||||||||
| 文献关键字 | |||||||||||
| encoder; decoder; content-based; cold-start; Symmetric recovery; projection; lowe-rank; sparsity | |||||||||||
| 摘要描述 | |||||||||||
| Zero-shot learning (ZSL) and cold-start recommendation (CSR) are two challenging problems in computer vision and recommender system, respectively. In general, they are inde- pendently investigated in different communities. This paper, however, reveals that ZSL and CSR are two extensions of the same intension. Both of them, for instance, attempt to predict unseen classes and involve two spaces, one for direct feature representation and the other for supplementary description. Yet there is no existing approach which addresses CSR from the ZSL perspective. This work, for the first time, formulates CSR as a ZSL problem, and a tailor-made ZSL method is proposed to handle CSR. Specifically, we propose a Low- rank Linear Auto-Encoder (LLAE), which challenges three cruxes, i.e., domain shift, spurious correlations and comput- ing efficiency, in this paper. LLAE consists of two parts, a low-rank encoder maps user behavior into user attributes and a symmetric decoder reconstructs user behavior from user at- tributes. Extensive experiments on both ZSL and CSR tasks verify that the proposed method is a win-win formulation, i.e., not only can CSR be handled by ZSL models with a signif- icant performance improvement compared with several con- ventional state-of-the-art methods, but the consideration of CSR can benefit ZSL as well. | |||||||||||
