Understanding RL for model training, and future directions with GRAPE

Don’t people google their newly coined acronyms? GRAPE is already Gradient-Ascent-Pulse-Engineering, which is arguably “machine learning” (optimal control)