A First Approach to Noise-Adaptive Accelerated Second-Order Methods

Over the last few decades, first-order (convex) minimization methods have gained popularity for modern machine learning and optimization problems due to their efficient per-iteration cost and global convergence properties.