From the course: Building Recommender Systems with Machine Learning and AI

Unlock this course with a free trial

Join today to access over 23,200 courses taught by industry experts.

Case study: YouTube, part 2

Case study: YouTube, part 2 - Python Tutorial

From the course: Building Recommender Systems with Machine Learning and AI

Start my 1-month free trial Buy for my team

Case study: YouTube, part 2

“

- [Instructor] YouTube was kind enough to publish their deep learning architecture, at least as it was in 2016. Let's start at the bottom, where we have the user behavior data that is used to train the system. It's interesting that, although YouTube has explicit ratings in the form of thumbs-up thumbs-down ratings, they don't use them at all for generating recommendations because that data is too sparse. Not enough users rate videos explicitly for the data to be useful. Instead, they rely on implicit signals, such as which videos you actually watched and what you searched for. This implicit view in search data, however, is in itself sparse. And as we learned, when covering deep learning recommenders, dealing with that sparsity is a huge issue when trying to apply deep learning to recommender systems. Their solution was to break up the sparse representation of video IDs and search tokens for each user into a variable length sequence of sparse data, mapped to a dense layer of a fixed…

Contents