Deep Learning

This seminar works through the foundations of deep learning by following modern textbooks. Students will each take ownership of a topic, present it to the group, and lead the subsequent discussion. The primary reference is Bishop’s “Deep Learning: Foundations and Concepts,” complemented by selected chapters from the textbook “Foundations of Computer Vision” (Torralba, Isola, and Freeman), specifically those on deep and representation learning, leaving aside the chapters dedicated to computer vision proper. Topics include the foundations of gradient-based learning and backpropagation, generalization, neural network architectures (convolutional networks, recurrent networks, and transformers), and an introduction to representation learning and generative models. The emphasis is on building a solid conceptual understanding of the core ideas rather than surveying the latest results, giving students a coherent and lasting foundation in the field.