Hardware Platforms for Deep Learning and Optimization

(3 Credit Hours) · Undergraduate

Modern deep learning algorithms have revolutionized machine learning. However, the computational complexity of deep learning hinders practical and real-time execution of such algorithms on many resource-limited devices. This course will cover custom digital platforms, analog chips, compute-in-memory architectures and implementations, as well as algorithmic techniques aimed at tailoring machine learning to hardware implementations. The course will include one project which will build upon the concepts learned over the semester. Grades will be based on class presentations, a final project, and class participation. There is no final exam for this course.