[Lecture] Building a Modern Machine Learning Accelerator - Outline

About Lecture

This was a short-term special course offered by Seoul National University (SNU).
The topic was “Building a Modern Machine Learning Accelerator”, and the goals of this short course are as follows:

1. Why accelerators for modern machine learning?
2. What are we accelerating and how?
3. What do modern accelerators look like, and why?

Instructor

• Sang-Woo Jun
  • Assistant Professor, UC Irvine
  • B.S. (2010) @SNU
  • Ph.D. (2018) @ MIT
• Research Interests
  • Systems architecture
  • Accelerators
  • NVM storage
  • Applications!
    • Graphs, Bioinformatics, Machine learning…
• ARDA Group @ UCI
  • ARchitecture for DAta

State of the World

Can you tell me the most important thing new researchers need to learn about building hardware and systems to enable future machine learning applications?

One of the most important things new researchers need to learn about building hardware and systems or future machine learning application is:

The critical importance of co-design between hardware, software, and algorithms.