Evidence item

A peer-reviewed Nature paper reports that AlphaTensor discovered provably correct matrix-multiplication algorithms outperforming prior state-of-the-art methods in several cases. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 9759

Source: Discovering faster matrix multiplication algorithms with reinforcement learning · locator {} · excerpt hash sha256:ba9ca8536f0d98d97633f8bea3415de5211cff1c53fa653bcc40c6942c48a802

Here we report a deep reinforcement learning approach based on AlphaZero for discovering efficient and provably correct algorithms for the multiplication of arbitrary matrices. Our agent, AlphaTensor, ... discovered algorithms that outperform the state-of-the-art complexity for many matrix sizes.