INTRODUCTION TO DEEP LEARNING

Credit Hours: 3 · Undergraduate Upper-Level

This course explores the design landscape of deep neural network architectures and optimization strategies, with the primary goal of giving students skills and knowledge that will help them as practitioners. After completing the course, students should be able to evaluate the tradeoffs of using different neural network building blocks and training strategies and understand how to choose the types of models that are better suited for a task. Specific topics covered include multi-layer perceptrons, backpropagation, convolutional neural networks, recurrent neural networks, autoregressive networks, and deep generative models. The notion of inner representation and embeddings as a semantic representation of inputs. Understand self-supervised vs supervised representation learning, including generative pre-training and brief introduction to multimodal representation learning. Case studies from applications such as computer vision and natural language processing will be used to illustrate the utility of various deep neural network designs and training strategies.