Artificial Intelligence on Chip

4 units · Graduate courses (200-299)

Lecture, four hours; outside study, eight hours. Recommended requisites: courses 115C, M116C, C147A. Introduction to hardware-aware machine learning (ML) with applications in computer vision, natural language, and big data. Students become better users and developers of artificial intelligence (AI) chips with software and hardware co-optimization, considering applications, algorithms, microarchitectures, circuits, and technologies for AI computing. Topics include neural network compression; operators, dataflows, and exemplar ML accelerator architectures; distributed training/inference; emerging computing models for AI. Example project topics include network compression, accelerator architecture modeling, hardware impact modeling of distributed inference. Letter grading.