youtube 代表了目前规模最大、最复杂的工业推荐系统之一。在这篇文章里，我们从系统的角度上重点讲述深度学习带来的巨大效果提升。根据经典的信息检索二分法，本文分为2阶段：首先，我们详细描述了一个候选集生成模型和一个深度排序模型。然后，我们还提供了一些从设计、迭代和维护庞大用户量的推荐系统中得到的实战经验和见解。

冷启动用户推荐，该文章发表于ICONIP-2020

Hands-On Recommendation Systems with Python: Start building powerful and personalized, recommendation engines with Python Rounak Banik 201807 Packt Publishing

My name is Frank Kane. I spent nine years at amazon.com and imdb.com, wrangling millions of customer ratings and customer transactions to produce things such as personalized recommendations for movies and products and "people who bought this also bought." I tell you, I wish we had Apache Spark back