How to understand the effectiveness of machine learning optimization algorithms, such as stochastic gradient descent, from the perspective of theoretical physics (link)
The connections between the well-known Kolmogorov-Arnold theorem, from real analysis, and the impressive generalization power of artificial neural networks (link)
The reasons why artificial neural networks can predict the outcomes of almost any process in nature (link).
How restricted Boltzmann machines (RBMs), building blocks of deep neural networks, can be used to compute the state of lowest energy of certain kinds of quantum systems (link).
A detailed description of the interconnections between deep learning and renormalization group theory (link).