Tag: per-instance RL

Per-instance TSP solver learns from scratch, hits 1.66% gap on d1291

TL;DR A developer posted a reinforcement-learning TSP solver that trains from scratch on a single instance and reports a 1.66%…