YouTube's mobile apps get a redesign and now use machine learning for better video recommendations


Posts: 7,435   +67
Staff member

YouTube has rolled out an update to its Android and iOS apps that sees the homepage get a redesign and introduces a new recommendation system based on deep neural network technology.

The way that the app grouped recommended items using small thumbnails has now been replaced with a single feed. This new list “features large, high-resolution images for identifying videos and an icon highlighting the creator of each video.”

In addition to the redesign, YouTube has promised that its new recommendation system will show items that are more relevant to users. It uses machine learning and algorithms to figure out your viewing patterns and improve the accuracy of recommendations the more videos you watch. It’s able to take into account videos that were uploaded within the last hour, as well as items from your YouTube subscriptions.

YouTube tested the new system on a small group of users in the weeks preceding the launch. The company found that they watched a larger number of videos and spent more time with them.

“We have made so many changes and we are getting so much better that we can now remove the grouping,” said YouTube’s VP of product management, Johanna Wright. “Our experiments show that users really like this design. We are a very data-driven group and all these changes run through A/B tests.”

While the visual updates are limited to the mobile incarnations of YouTube, the new recommendation technology will also be coming to the desktop version.

It was reported yesterday that YouTube will be rolling out six-second, unskippable ‘Bumper ads’ for mobile devices next month.

Permalink to story.