How does the Youtube Video Recommendation work?

[Brief Overview]

9 June, 2022

Contributors

Rohan Fulzele

@rohn-dev

Hey Developer,🙋🏻‍♂️

Have you ever wondered how you can go to YouTube looking for a video on a specific topic and wind up spending hours of your day just viewing the recommended videos, or how you can go to YouTube looking for your favorite music track and end up adding it to your favorites list?

Many content-driven businesses are pouring money into making their recommender systems as good as or better than YouTube's, in order to make the site more interesting for us.

✨ let's see how,

✅To make video recommendations, the YouTube recommendation engine employs two neural networks and a variety of data sources.

The Candidate Generation Network and the Ranking Network are the two networks that make up the YouTube recommendation engine.

✅The Candidate Generation Network retrieves several hundred videos from the YouTube video archive based on the user's history derived from log data and the video context. A cooking presentation, for example, is not combined with a movie of football highlights.

✅These hundreds of videos are supplied to the Ranking Network, which references and uses videos from sources other than the user's log data, such as recent uploads to show fresh and likely unwatched videos, sponsored films, and news videos, among others.

✅These data sources from the Candidate Generation Network and the Ranking Network are then combined to reduce the number of videos that can be recommended to only a few dozen.

✅However, training this hybrid model is more complicated than just integrating the results of two massive networks. Each has its own complex architecture that must be taken into account.

If you want to dive deeper into this...please check out the link below -