This course aims to present the mathematical, and computational challenges for holistic content/algorithm/hardware co-design of an efficient Deep Learning (DL) framework. We will discuss selected DL topics including DNNs, CNNs, GNN and Transformers in both supervised and unsupervised settings. Special emphasis will be on optimizing DL physical performance (e.g., real-time performance, energy, memory, and power) on different hardware platforms.

  • Instructor: Prof. Farinaz Koushanfar, UCSD ECE
  • Time: Tues/Thurs, 12:30-13:50 PM PT
  • Location: Cognitive Science Building 002
  • Office Hours: Thurs 2:00-3:30 PM, Location EBU1 #6104 (Or by email appointment)
  • Discussion: Piazza
  • Important Notice: Please note that all sessions of this class will be podcasted and made available to students asynchronously.