Gradient Descent with Provably Tuned Learning-rate Schedules

Dravyansh Sharma

Gradient-based iterative optimization methods are the workhorse of modern machine learning. They crucially rely on careful tuning of parameters like learning rate and momentum. How…