【阅】本周阅读摘选2026-01-12 → 2026-01-18

Posted by Cao Zihang on January 19, 2026 Word Count:
本周阅读摘选
2026-01-12 → 2026-01-18
目录

学术相关

Revenue Maximization and Learning in Product Ranking 1

Learning user preferences in livestreaming market: A graphical model considering temporal effect 2

Matrix Factorization (补充)

在推荐系统中,用户与商品之间的交互数据(如评分、点击、购买等)通常表示为$m\times n$的系数矩阵$R$

  • $m$ 为用户数量
  • $n$ 为商品数量
  • $R_{i,j}$ 表示用户$i$对商品$j$的交互数据

矩阵分解的核心目的是将高维系数矩阵$R$分解为两个低维稠密矩阵的乘积,从而挖掘交互的潜在特征

$R\approx P\times Q^T$

  • $P \in \mathbb{R}^{m \times k}$ 为用户隐因子矩阵,每行$p_u$表示用户$u$的隐因子向量
  • $Q \in \mathbb{R}^{n \times k}$ 为商品隐因子矩阵,每列$q_i$表示商品$i$的隐因子向量
  • $k$ 为隐因子的维数,通常远小于$m$和$n$
  • $r_{ui}\approx p_u^Tq_i$ 为用户$u$对商品$i$的评分$

该方法擅长处理(适度)稀疏矩阵,常用于显示评分预测,隐式反馈推荐,具有较强的泛化能力和可解释性;缺点是存在冷启动问题,计算复杂度高,超参数敏感

Replicator dynamics and behavior-augmented multiscale epidemic modeling 3

技术技巧

Learning Prompting

一个比较全面的提示词学习网站

  1. Chen, N., Li, A., & Yang, S. (2026). Revenue Maximization and Learning in Product Ranking. Operations Research, opre.2020.781. https://doi.org/10.1287/opre.2020.0781 

  2. Lin, Q., Li, Y., Kadziński, M., & Guo, M. (2026). Learning user preferences in livestreaming market: A graphical model considering temporal effect. Decision Support Systems, 202, 114600. https://doi.org/10.1016/j.dss.2025.114600 

  3. Wan, J., Liu, Q., Zacks, K., Zhao, J., Ichinose, G., Small, M., & Cheng, C. (2025). Replicator dynamics and behavior-augmented multiscale epidemic modeling. Chaos: An Interdisciplinary Journal of Nonlinear Science, 35(12), 123128. https://doi.org/10.1063/5.0297383