Buffio.

A field journal of machine-made science

← All discoveries

Mathematics · 14 May 2025

A faster algorithm for 4×4 complex matrix multiplication

AlphaEvolve found an algorithm that multiplies 4×4 complex-valued matrices using 48 scalar multiplications, improving on Strassen’s 1969 algorithm, which uses 49 in that setting.

Field
Mathematics
Date
14 May 2025
AI system
AlphaEvolve
Institution
Google DeepMind
Published in
Official announcement and technical report
Identifier
See primary source

AlphaEvolve is an evolutionary coding agent, announced by Google DeepMind on 14 May 2025, that pairs Gemini models with automated evaluators to propose and improve algorithms. Unlike a system that emits a single function, it can evolve larger programs.

In matrix multiplication, a fundamental operation in scientific computing, it found an algorithm to multiply 4×4 complex-valued matrices using 48 scalar multiplications. That improves on Strassen’s 1969 algorithm, which uses 49 multiplications in this setting. DeepMind notes that this goes beyond AlphaTensor, which had found improvements for 4×4 matrices only in binary arithmetic.

The same announcement reports a second mathematical result: on the kissing-number problem — the maximum number of non-overlapping spheres that can touch a central unit sphere — AlphaEvolve discovered a configuration of 593 outer spheres and established a new lower bound in 11 dimensions. Applied to more than 50 open problems in analysis, geometry, combinatorics and number theory, the system rediscovered state-of-the-art solutions in roughly 75% of cases and improved the previously best-known solutions in 20% of cases, according to the official account.