Deep Learning
COMPSCI 2850
Subject & Catalog Number
Course Information
Description
This graduate course introduces the principles and practice of deep learning, from neural network training to foundation models and agents. Topics include automatic differentiation, optimization, convolutional and sequence models, transformers, generative modeling, data and scaling, distributed training, adaptation and post-training, efficient inference, and reasoning with tools and memory. Applications in language, vision, robotics, and biology illustrate how modeling choices, learning objectives, and computation shape capabilities. Programming assignments develop practical understanding, while a midterm and final examination assess mastery of the underlying principles.
Recommended Prep
Prior machine learning at the level of CS1810 or equivalent; linear algebra, multivariable calculus, basic probability, and proficiency in Python. Experience with a tensor library is helpful but not required.
Available for Harvard Cross Registration
