Read on Twitter
Today, we're describing the ML models which power the ranking system personalizing the News Feed of every person around the world who uses Facebook.
Here's how ranking works: https://tech.fb.com/news-feed-ranking/
And here is the engineering behind it: https://engineering.fb.com/2021/01/26/ml-applications/news-feed-ranking/
Here's how ranking works: https://tech.fb.com/news-feed-ranking/
And here is the engineering behind it: https://engineering.fb.com/2021/01/26/ml-applications/news-feed-ranking/
When it comes to ranking content in people’s News Feeds, our number one goal is to connect people with the posts they will find most meaningful and we’re always improving our products to do just that: https://tech.fb.com/news-feed-ranking/
To start, the ranking system collects all the candidate posts we can possibly rank for a given user, including any post shared by a friend, Group, or Page you’re connected to that was made since your last login and hasn’t been deleted: https://tech.fb.com/news-feed-ranking/
Then, the system scores each post based on a variety of factors, such as whether you’ve told us a post is worth your time, who posted it, and how much the post matches what you tend to interact with. Integrity processes are applied to every post: https://tech.fb.com/news-feed-ranking/
In the next pass, a lightweight model narrows the pool of candidates to the most relevant posts. Ranking fewer stories allows us to use more powerful neural network models for the next passes: https://tech.fb.com/news-feed-ranking/
Next is the main scoring pass, where a score for each story is calculated independently and then all posts ordered by score. Finally, we run the contextual pass, in which contextual features are added to make sure your News Feed has a good mix of content: https://tech.fb.com/news-feed-ranking/
The News Feed ranking system is not just a single algorithm; it’s multiple layers of ML models that we apply to predict the content that’s most relevant and meaningful for each user.
We explain those ML models here: https://engineering.fb.com/2021/01/26/ml-applications/news-feed-ranking/
We explain those ML models here: https://engineering.fb.com/2021/01/26/ml-applications/news-feed-ranking/
The science and engineering expertise that drives the News Feed ranking architecture is designed to create a valuable experience for people at remarkable scale and speed: https://engineering.fb.com/2021/01/26/ml-applications/news-feed-ranking/
