Seminar on Hardware Accelerators for AI (IN0014, IN2107)
Dirk Stober (TUM), Prof. Miloš Krstić (IHP, UP), Dedong Zhao (UP)
| Dates: | Online Lecture: Fridays 10:00-12:00 (zoom,recorded) Presentation (zoom): Wed. 10.07 (16:00-18:00), Fri. 12.07 (14:00-16:00) and Wed 17.07 (16:00-18:00) |
| Kick-Off Meeting: | 29.04.2025 (16:00-17:00) (zoom) |
| ECTS: | 5 |
| Language: | English |
| Type: | Bachelor/Master Seminar course |
| Moodle course: | openUP moodle (external) |
| Registration: | Registration is through the matching system |
| Questions? | Contact dirk.stober(at)tum.de |
This course is part of the BB-KI (Brandenburg / Bayern Aktion für KI-Hardware) chips project, aimed at offering practical courses in the area of dedicated AI Hardware.
Topics
- Introduction in VLSI and digital logic design
- Hardware Design Process, ASIC, FPGA
- Von Neumann Computing Architecture
- State of the art processor architecture, Example RISC-V
- Limitations of classical architectures for AI applications
- Accelerators architectures: GPUs, MAC arrays
- Neuromorphic Architectures (TrueNorth, Loihi, Google architectures)
- Emerging architectures: In-Memory-Computing (example RRAM)
Organization
This course consists of a weekly lecture introducing concepts in HW accelerator design with a specific focus on AI. This will prepare the students to perform their own literature survey for a selected paper. The course is taught in cooperation with the University of Potsdam and lectures are held remote and will be available as videos as well. To pass the course students have to answer at least 50% of the weekly moodle questions concerning the lecture. In addition students have to hand-in a report (Literature survey) on a selected topic and present their topic in a 15 min presentation in the last few weeks of the semester. The course material will be available on the UP moodle page.
Kick off Meeting:
- Presentation of course organization + report topics
- 29.04.2025 16:00-17:00 (online)
Lectures (Friday 10:00-12:00) & Weekly Questions:
- Online or pre-recorded
- Weekly Questions on open.UP moodle (50% to pass course)
Report:
- Literary survey
- Concepts and Trade-offs
- ~15 min presentation + Questions (Last Weeks of Semester)
- Deadline: End of semester
Grading:
The work will be performed on an individual basis and the final grade will be based on the sum of the three grades (Presentation , Report and Weekly Questions), with all three tasks being mandatory to pass.
Prerequisites:
Basic knowledge in Computer Science.