Scion is a new stochastic LMO-based optimizer family that unifies existing methods, supports unconstrained problems, and delivers hyperparameter transferability plus speedups on nanoGPT training.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.LG 2verdicts
UNVERDICTED 2representative citing papers
GONO adapts Adam's momentum using measured gradient directional consistency to better navigate plateaus and oscillations while matching Adam's convergence rate.
citing papers explorer
-
Training Deep Learning Models with Norm-Constrained LMOs
Scion is a new stochastic LMO-based optimizer family that unifies existing methods, supports unconstrained problems, and delivers hyperparameter transferability plus speedups on nanoGPT training.
-
Directional Consistency as a Complementary Optimization Signal: The GONO Framework
GONO adapts Adam's momentum using measured gradient directional consistency to better navigate plateaus and oscillations while matching Adam's convergence rate.