TY - RPRT TI - Learning Rate Decay Can Exponentially Accelerate SGD for Global Optimization of Nonconvex Functions AU - Guneykan Ozgul AU - Dylan Herman AU - Junhyung Lyle Kim AU - Shree Hari Sureshbabu AU - Shouvanik Chakrabarti PY - 2026 UR - https://searcharxiv.com/abs/2611.12769 ID - 2611.12769 N1 - Source identifier: 2611.12769 ER -