QAC 388 – Deep Learning and Large Language Models: An Introduction

You have probably used ChatGPT, DeepSeek, Midjourney, or another generative AI in your lifetime. But how do these tools actually work? This course will help you answer that question by introducing you to the theory and practice of deep learning, the fundamental technology underlying generative AI. Students will learn the building blocks of modern neural networks, including matrix and vector operations, embeddings, backpropagation, and attention. This will culminate in the implementation of the original GPT transformer language model from scratch in PyTorch. We will then explore extensions of transformers in other modalities, including vision and diffusion transformers for understanding and generating images, and transformers for time series analysis. Along the way, we will cover other types of neural networks, including multilayer perceptrons, CNNs, and RNNs. We will use these models in a variety of practical applications using real-world datasets.