Gradient Descent More Efficient with Periodic Long Steps (For Convex Functions) | Hacker News Reader