Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Volume 610 Issue 7930, 6 October 2022

Matrix games

The cover shows an artistic impression of a matrix multiplication tensor — a 3D array of numbers — in the process of being solved by deep learning. Efficient matrix multiplication algorithms can help speed up many computations, and in this week’s issue, Alhussein Fawzi and his colleagues at DeepMind show how machine learning can uncover faster algorithms. The system, called AlphaTensor, was trained on a game that involved finding the best way to ‘decompose’ a matrix multiplication tensor so as to find matrix multiplication algorithms. After training, AlphaTensor was able to rediscover previously known algorithms as well as to uncover new ones that, in some cases, improved on algorithms that have resisted improvement for more than 50 years.

Cover image: Adam Cain/Domhnall Malone/DeepMind

This Week

Top of page ⤴

News in Focus

Top of page ⤴

Opinion

Top of page ⤴

Work

Top of page ⤴

Research

Top of page ⤴

Amendments & Corrections

Top of page ⤴

Collections

  • At some institutions around the world, researchers work hand-in-hand with medical professionals to the benefit of both.

    Spotlight
Top of page ⤴
Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing

Search

Quick links