Understanding Machine Learning: Exactly Solvable Models

This lecture by the instructor Lenka Zdeborová delves into the statistical mechanics of learning, exploring the mysteries behind neural networks. Topics covered include sample complexity, models in data science and physics, and the learning capabilities of neural networks. The lecture also discusses the teacher-student perceptron, the optimal storage capacity of networks, and the generalization error in neural networks. Through various models and algorithms, the lecture aims to shed light on the theoretical aspects of deep learning and the computational challenges in avoiding spurious minima.