AI 相关¶
推荐系统¶
1. 内容推荐¶
[论文, Facebook]Bag of Tricks for Efficient Text Classification:
Facebook 开源的文本处理工具 fastText 背后原理。 可以训练词嵌入向量,文本多分类,效率和线性模型一样,效果和深度学习一样,值得拥有。
[论文, Google]The Learning Behind Gmail Priority Inbox:
介绍了一种基于文本和行为给用户建模的思路,是信息流推荐的早期探索,Gmail 智能邮箱背后的原理。
Recommender Systems Handbook:
作者:Francesco Ricci 等 这本书收录了推荐系统很多经典论文,话题涵盖非常广, 第三章专门讲内容推荐的基本原理, 第九章是一个具体的基于内容推荐系统的案例。
2. 近邻推荐¶
[论文, Amazon]Amazon.com recommendations: item-to-item collaborative filtering:
介绍 Amazon 的推荐系统原理,主要是介绍 Item-Based 协同过滤算法。
[论文, Daniel Lemire]Slope One Predictors for Online Rating-Based Collaborative Filtering:
Slope One 算法。
[论文, Badrul Sarwar]Item-Based Collaborative Filtering Recommendation Algorithms:
GroupLens 的研究团队对比了不同的 Item-to-Item 的推荐算法。
3. 矩阵分解¶
[论文, Yehuda Koren]Factorization Meets the Neighborhood: a Multifaceted Collaborative Filtering Model:
把矩阵分解和近邻模型融合在一起。
[论文, Steffen Rendle]BPR- Bayesian Personalized Ranking from Implicit Feedback:
更关注推荐结果的排序好坏,而不是评分预测精度,那么 BPR 模型可能是首选,本篇是出处。
[论文, Yifan Hu]Collaborative Filtering for Implicit Feedback Datasets:
不同于通常矩阵分解处理的都是评分数据这样的显式反馈,本文介绍一种处理点击等隐式反馈数据的矩阵分解模型。
[论文, Yehuda Koren]Matrix Factorization Techniques For Recommender Systems:
本文是大神 Yehuda Koren 对矩阵分解在推荐系统中的应用做的一个普及性介绍,值得一读。
[论文, Yehuda Koren]The BellKor Solution to the Netflix Grand Prize
4. 模型融合¶
Adaptive Bound Optimization for Online Convex Optimization
Ad Click Prediction: a View from the Trenches
Factorization Machines
Field-aware Factorization Machines for CTR Prediction
Practical Lessons from Predicting Clicks on Ads at Facebook
Wide & Deep Learning for Recommender Systems
5. Bandit 算法¶
A Contextual-Bandit Approach to Personalized News Article Recommendation
Collaborative Filtering Bandits
6. 深度学习¶
Deep Neural Networks for YouTube Recommendations
Efficient Estimation of Word Representations in Vector Space
Item2Vec: Neural Item Embedding for Collaborative Filtering
Learning Representations of Text using Neural Networks
Long Short-Term Memory
An Empirical Exploration of Recurrent Network Architectures
7. 其他实用算法¶
Detecting Near-Duplicates for Web Crawling
Weighted Random Sampling over Data Streams
Weighted Sampling Without Replacement from Data Streams
工程篇¶
Information Seeking-Convergence of Search, Recommendations and Advertising
Overlapping Experiment Infrastructure- More, Better, Faster Experimentation
TencentRec:Real-time Stream Recommendation in Practice