Tony Geng

Machine learning is in tremendous demand across a wide range of applications. However, the recent extreme scaling of machine learning models has introduced prohibitive computational complexity, creating a major barrier to their broad deployment in real-world and resource-constrained environments. This course introduces techniques for developing energy- and time-efficient machine learning systems through a comprehensive pathway spanning algorithms, computer architecture, and their co-design. Students will first learn commonly used machine learning algorithms, modern computer architecture paradigms, and the principles underlying their intersection and co-optimization to reduce energy and latency while preserving strong model performance. The course then extends beyond current mainstream models to explore emerging learning paradigms, including brain-inspired artificial intelligence and thermodynamics-inspired computing, offering a forward-looking perspective on next-generation machine learning systems. Cross-list: ELEC 515. Recommended Prerequisite(s): Students should be familiar with PyTorch programming, AI basics, computer architecture, and digital circuit design to complete the course projects. Mutually Exclusive: Cannot register for ELEC 415 if student has credit for ELEC 515.

Machine learning is in tremendous demand across a wide range of applications. However, the recent extreme scaling of machine learning models has introduced prohibitive computational complexity, creating a major barrier to their broad deployment in real-world and resource-constrained environments. This course introduces techniques for developing energy- and time-efficient machine learning systems through a comprehensive pathway spanning algorithms, computer architecture, and their co-design. Students will first learn commonly used machine learning algorithms, modern computer architecture paradigms, and the principles underlying their intersection and co-optimization to reduce energy and latency while preserving strong model performance. The course then extends beyond current mainstream models to explore emerging learning paradigms, including brain-inspired artificial intelligence and thermodynamics-inspired computing, offering a forward-looking perspective on next-generation machine learning systems. Additional course work required beyond the undergraduate course requirement. Cross-list: ELEC 415. Mutually Exclusive: Cannot register for ELEC 515 if student has credit for ELEC 415. Cross-list: ELEC 415. Mutually Exclusive: Cannot register for ELEC 515 if student has credit for ELEC 415.

Repeatable for Credit.