I Physical View of Deep Learning
15
In Part I, the key point is how to model the mathematics of learning as a physical
model and, as a result, obtain the fact that the neural network is derived physically
and naturally. These will be useful for various occasions: for further learning
about various constructions of machine learning architecture, for understanding new
learning methods that will show up in the future, for applying physics to the research
of machine learning, and for applying machine learning to the study of physics.