This is Lecture 17 of the course on Probabilistic Machine Learning in the Summer Term of 2025 at the University of Tübingen, taught by Prof. Philipp Hennig.
Contents include an explicit identification of general deep learning architectures (empirical risk minimization for differentiable programs) with an approximate Gaussian process posterior through the linearized Laplace approximation.
Probabilistic ML is an integral part of the curriculum of the International Masters Degree in Machine Learning, alongside associated courses on deep learning, statistical machine learning, reinforcement learning, and much more.
Playlist for the course: https://youtube.com/playlist?list=PL05umP7R6ij0hPfU7Yuz8J9WXjlb3MFjm&feature=shared
