《唯快不破：推荐系统时效性的实践与思考-陈天健》由会员分享，可在线阅读，更多相关《唯快不破：推荐系统时效性的实践与思考-陈天健（23页珍藏版）》请在金锄头文库上搜索。
1、 InfoQ infoqchina 唯快不破 推荐系统时效性的实践与思考 Tianjian Chen Baidu Inc. About Me Principal Architect Baidu Inc. C/C+/Python Programmer RecSys Designer Engineer Psychological Trauma Therapist 推荐系统概要 RecSys Overview Algorithm Family Ranking Optimizer + Merge Ranking Algorithms Neighborhood Based Methods Content
2、 Based Methods Constraint Based Methods Context- Aware Methods Neighborhood Based Recommendation X Based Collaborative Filtering X can be user, item, model, etc. Content Based Recommendation Mr. John Doe likes “Cars” “Ferrari” is a car John Doe may like “Ferrari” Constraint Based Recommendation 为什么要
3、变快 Why RecSys needs to be faster? Technology Section More Accurate User Profile More Accurate Ranking Model CTR Model Time-Efficiency Product Section 1st Scenario Product Section 1st Scenario 2nd Item Quality vs. Product Section 1st Scenario 2nd Item Quality 3rd Time-Efficiency 如何变快 How to accelerate your RS? Stream computing makes your RS fast and flexible Type Based C-F Mercedes BMW Audi Toyota Honda Ford GM Thank U! Email: 特别感谢合作伙伴 特别感谢媒体伙伴（部分）
