


DEPARTMENT OF STATISTICAL SCIENCE DUKE UNIVERSITY


Statistics 790 (Duke Numbering) Professor David Banks This course is part of the Statistical and Applied Mathematical Sciences Institute's program on Deep Learning, and admission is restricted. The course will cover the theory and practice of deep learning. It will begin with a review of current applications of deep neural networks, then discuss the basics, including the perceptron, backpropagation, and the universal approximation theorem. From there it will proceed to boosted coordinate gradient descent, convolutional neural networks, recurrent and recursive neural networks, and generative adversarial networks. Students will work in teams of size three on six classic challenge datasets. Please ensure that at least one member of each team has some familiarity with Python or some other programming language. The course will not have exams, but there will be regular applicationsoriented homework. 