LinkedIn improves news feed relevance using artificial intelligence

LinkedIn just announced significant progress in using artificial intelligence to improve the relevance of its news feed. By reporting the latest developments in this area, LinkedIn is positioned to provide a more personalized and engaging experience for its members. Find out in this article the content of this last statement.

What AI will bring to LinkedIn in #Short

  1. The use of artificial intelligence is essential to improve the relevance of the news feed on LinkedIn.
  2. LinkedIn uses ” incorporations to transform sparse identifiers into embedding spaces, which help capture important relationships and patterns across data while reducing computational complexity.
  3. LinkedIn has significantly has increased the size of its modelsallowing for more precise customization and better consideration of complex interactions.
  4. The use of model parallelization has greatly improved training efficiency by allowing models containing billions of parameters to be trained in a reasonable amount of time.
  5. LinkedIn migrated from an external service model to an in-memory serviceallowing for faster feature deployment and more flexibility for modelers.

Here’s how LinkedIn invests in AI on its platform

Replying to his pursuit of continuous improvement, LinkedIn recently made significant improvements to its architecture. These improvements are designed to optimize process efficiency while maintaining outstanding performance.

This modernization aims to increase the quality of service and facilitate the implementation of other large-scale innovations within the platform. We will therefore focus on it in this article.

The role of artificial intelligence in this search for improvement

AI is at the heart of LinkedIn’s News Feed relevance strategy. Leveraging deep learning, the platform explore very large datasets deposit:

  • Discover complex patterns
  • Identify relevant relationships
  • Deliver more targeted and meaningful content to every member

The Icustom incorporations

To turn infrequent identifiers into embedding spaces, LinkedIn has taken an approach that can only be described as innovative.

These embed spaces capture essential information, including each member’s preferences and past interactions.

This technique allows you to simplify the calculations while retaining the wealth of information. Result? More developed personalization and more precise recommendations.

Through the use of these personalized embeds, LinkedIn is able to recommend relevant and engaging content to each member. Based on past interactions, preferences, and professional relationships, the News Feed becomes an invaluable source of content that caters to each user’s unique needs.

Adoption of the parallelization model and modification of the service infrastructure

By parallelizing learning patterns, LinkedIn has improved the efficiency of its model training. This approach significantly reduced the time required to train models consisting of billions of parameters, while ensuring high-quality results.

LinkedIn has also made changes to its infrastructure, evolving from an external service model to an in-memory service model, which allowed features to be delivered more quickly and with greater flexibility. This transition has not only improved operational efficiency, but also strengthened the pattern makers’ ability to innovate!

Bonus: Fixed “Cold Start” issue.

One of the main challenges LinkedIn faces is the “cold start” – the difficulty of recommending relevant content to new members or recently added items.

With these changes, and in particular with custom embeds, LinkedIn has found a solution to this problem by associating the new elements with the existing embed spaces, which helps to establish relationships and similarities even with little information.

Advantages of large models

The significant expansion of model sizes within the LinkedIn ecosystem has played a role pivotal role in improving the relevance of content offered to members. This increase in the ability of models to handle huge volumes of data and capture complex information has had a negative impact direct impact on customization offered to users.

By broadening their reach, the models are now able to analyze more subtle interactions, deeper patterns and nuances that would otherwise escape less extensive modeling. This finer and more detailed customization has therefore logically led to a significant improvement in the quality of the recommendations provided to members.

From now on, LinkedIn News Feed can better capture each user’s specific interests, preferences, and work habits.

Before leaving…

If this article is active the added value of artificial intelligence for LinkedIn you liked it, do not hesitate to share it on social networks ea subscribe to our digital newsletter to receive our next articles.

You can also follow our best articles via our RSS feed: (just insert it in your favorite RSS feed reader (ex: Feedly)).

We are also active on LinkedIn, Twitter, Facebook and YouTube.

For any question relating to this article, do not hesitate to use the “comments” section to send us your comments, we will reply (with pleasure) as soon as possible.


You wish to receive Our best articles ?

(You can cancel at any time.)

Source link

Related Posts

Leave a Reply

Open chat
Scan the code
Hello 👋
Can we help you?