Discovering Lattice Reduction Strategies via Self-Play
Researchers have developed a novel approach to lattice reduction strategies by employing deep reinforcement learning, specifically an AlphaZero-style self-play pipeline with Monte Carlo Tree Search. This method trains a deep residual network to discover strategies that outperform the traditional Lenstra-Lenstra-Lovász (LLL) algorithm. The resulting policy, DeltaStar, trained on small lattices, demonstrates generalization to higher dimensions and unseen moduli without retraining. AI
IMPACT AI-driven discovery of superior mathematical algorithms could accelerate progress in fields reliant on complex computations.