Science ❯ Computer Science ❯ Data Structures ❯ Graph Theory
The authors report up to 450x faster training with accuracy maintained even at D=128.