PILOT: Policy-Informed Learned Optimization for Adaptive Deep Network Training

ArXi:2605.24570v1 Announce Type: cross Despite the central role of optimization in deep learning, most optimizers rely on update structures whose functional form is fixed before