Please explain how Hopfield network influenced modern deep learning models based on supervised differentiable training. All the "impactful" architectures such as MLP, CNN, Attention, come from a completely different paradigm, a paradigm that could be more straightforwardly connected to optimization theory.