Distributed Newton-Like Algorithms And Learning For Optimized Power Dispatch