Neural Networks. 3 credits, 3 contact hours

Catalog description

Prerequisites: CS 375 with a grade of C or better. This course introduces modern neural network architectures, including multilayer perceptrons (MLPs), convolutional networks, recurrent networks, and transformers. It emphasizes the mathematical foundations of these models and their implementation in PyTorch, with attention to architectural design choices and training tradeoffs. Topics include optimization and regularization, sequence modeling, self-attention, transformer architectures, fine-tuning pretrained language models, and techniques for diagnosing and debugging training failures.