Speaker: Dmitry Rybin CUHK(SZ)
Time: 2025-05-29 10:00-2025-05-29 12:00
Venue: Tencent Meeting: 299-679-719 (Tencent Meeting: 299-679-719)

Abstract:

We present a new algorithm RXTX that computes product of matrix by its transpose XX^T for any size of X. RXTX uses 5% fewer additions and multiplications than other algorithms. Algorithm was discovered through combination of Machine Learning-based Search and Combinatorial Optimization. We discuss theoretical and practical implications, connections to AlphaEvolve and AlphaTensor.

Short Bio:

Dmitry Rybin is a 4th year Machine Learning PhD student at CUHK(SZ), under supervision of prof. Luo Zhi-Quan. He studied pure math at HSE, Moscow. Dmitry is a winner of International Math Olympiads (IMC, Alibaba Global Math, IZhO), and Russian National Math Olympiad.