Neural Networks and Quantum Field Theory

This lecture explores the connection between neural networks and quantum field theory, focusing on the correspondence between parameter and function spaces, symmetries via duality, and the concept of 'IR' dualities. It delves into the asymptotic behavior of neural networks, Gaussian processes, and free field theory, highlighting the interplay between NN distributions and QFT techniques. The lecture also discusses the implications of scale-invariance, correlation functions, and the existence of universality classes in neural networks. Through examples and experimental results, the instructor demonstrates how NN architectures can be understood through the lens of QFT, emphasizing the importance of symmetries and effective actions.