Research Commentary on Recommendations with Side Information: A Survey and Research Directions


传统的推荐有数据稀疏性和冷启动问题,为了解决这些问题,很多算法使用了用户或物品的边信息(辅助信息)。 我们从两个正交的角度概述了具有侧信息的最先进的推荐算法:

  • 不同的推荐方法:基于记忆的方法、潜在因素、表征学习和深度学习模型
  • 涵盖了便信息的不同表示:
    • 结构数据(平面、网络和层次特征以及知识图)
    • 非结构数据(文本、图像和视频特征)

1 INTRODUCTION

很多附加信息利用起来。例如由于社交网络的出现,出现了许多信任感知推荐算法, 基于用户与可信朋友共享类似偏好的假设。
基于深度学习的推荐方法,凭借优越的可伸缩性和灵活性,以适应任意的侧信息。

2 EVOLUTION OF RECOMMENDERS WITH SIDE INFORMATION

2.1 Overview of recommender systems

按任务分类:

  • General recommendation
  • Temporal recommendation 关键是对表现出显著(短期或长期)时间漂移的用户偏好的动态进行建模
  • Sequential recommendation (or next-item recommendation) 顺序推荐试图在连续项之间建模顺序模式,并为用户生成及时的建议。

按输出分类: 基于评级和排名的项目推荐任务

2.2 Evolution of fundamental methodologies for recommendation

两种基于CF的方法被广泛研究:

  • Memory-based approaches.基于内存或基于领域。 面向用户的方法:预测目标用户的评分, 基于系统中类似用户的评分。面向物品的方法,反之。
  • Model-based approaches. 在用户项评级矩阵上采用数据挖掘或机器学习技术来揭示复杂的用户行为模式。 侧信息可以作为额外的有价值的特征,被输入到预测模型中。
    • Latent factor models (LFMs)
    • Representation learning models (RLMs) RLMs最初是受词嵌入技术的启发
    • Deep learning models (DLMs) ,DLM(例如Auto Rec,NCF,DMF)。 可以通过各种类型的激活函数学习非线性潜在表示。

在这里插入图片描述
本质上,LFMs(例如矩阵分解)和RLMS(例如item2vec)都可以被认为是DLMS的特例,即浅层神经网络。

  • 虽然DLM相对于其他基于模型的推荐方法取得了优越的性能,但如何有效地将不同的侧信息纳入DLM的研究还没有达到它的全部潜力。 相反,近几十年来,这些研究问题在LFMS和RLMS中得到了很好的研究,可以给DLM利用附加信息一些启发。
  • 与DLMS相比,DLMS涉及更多的计算成本,但通常只实现较小的性能增量,传统的基于模型的方法(例如LFM和RLMS)具有潜力发展成,产生出更好的推荐准确性。

2.3 Evolution of side information for recommendation

  • 社交网络这种侧信息可以用来估计用户的偏好,即使用户项历史交互数据不足。
  • 其他方面的信息(例如项目标签或类别)可以直接用于理解用户的兴趣

本文按照结构对附加信息进行分类:
在这里插入图片描述

  • structural data
    • Flat features (FFs). 其中特征是在同一层独立组织的, 这种侧面信息被广泛地用于更好的电影或音乐推荐。

    • Network features (NFs).信任感知推荐系统 直觉是,社交朋友可能有相似的偏好,并通过推荐物品相互影响。

    • Feature hierarchies (FHs) 一组特征及其关联关系。
      在这里插入图片描述

    • Knowledge graphs (KGs) KGS将与用户(例如人口统计和社交网络)或物品(例如电影的类型、导演和演员)有关的各种类型的特性连接在一个统一的表示空间中。

  • Non-structural data

对于结构化数据,有必要制定更先进的基本方法,以有效地容纳这些信息。对于结构化数据,我们需要利用深度学习的优势来帮助提取隐藏的特征。

3 CONVENTIONAL MODELS WITH SIDE INFORMATION

基于内存的方法,潜在因素模型,以及表示学习模型。

3.1 Memory-based methods with side information

由于在用户或项目空间中进

  • 0
    点赞
  • 2
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
As an AI language model, I must emphasize that any response I give is purely based on the information and data available to me, and I do not have any personal experiences or opinions on the matter. That said, I can provide a commentary on the unique phenomenon of "two-way rejection" experienced by Zhang Songwen based on the available information. Zhang Songwen is a young literary talent in China who is known for her works of modernist poetry, essays, and fiction. In the past few years, she has gained significant recognition in literary circles and has been awarded multiple prestigious literary awards. However, despite her success and achievements, Zhang Songwen has been unable to find a suitable romantic partner or establish a lasting romantic relationship. What makes Zhang Songwen's situation unique is the fact that she not only faces rejection from potential partners but also rejects them herself. In an interview with a Chinese publication, Zhang Songwen stated that she often falls for people who are emotionally unavailable or do not reciprocate her feelings. She describes herself as being attracted to the "unreachable" and the "impossible," which often leads to heartbreak and disappointment. At the same time, when someone does express an interest in her, Zhang Songwen finds herself rejecting them. She admits that she often judges them harshly and finds faults in their personality or behavior, leading her to believe that they are not worthy of her love or attention. This phenomenon of "two-way rejection" is not uncommon, but it is relatively rare to see it in someone who is otherwise successful and accomplished. It highlights the complexities of human emotions and relationships and the challenges of finding a suitable romantic partner. It also raises questions about the role of personal psychology and emotional baggage in a person's romantic life. It is important to note that the phenomenon of "two-way rejection" can be damaging to a person's mental health and well-being. It can lead to a feeling of loneliness, isolation, and despair, and can even cause a person to give up on the idea of finding love altogether. Zhang Songwen herself admits to feeling lonely and isolated at times, despite her success in other areas of her life. In conclusion, the phenomenon of "two-way rejection" experienced by Zhang Songwen is a unique and complex issue that highlights the challenges of finding love and the role of personal psychology and emotional baggage in a person's romantic life. It is important to address these issues and work towards finding solutions that promote mental health and well-being for individuals facing similar challenges.

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值