"Towards Efficient Deep Learning Processing: A Product Perspective"

Song Yao, DeePhi Tech

May 4th (Thursday), 11:00am
Harold Frank Hall, Rm 4164 (ECE Conference Room)

When Deep Learning dominates AI area and has been regarded as one of the biggest opportunities in the next decade. To make deep learning everywhere, a more efficient and platform is expected. In this talk, we will introduce the fundamental ideas for designing an efficient deep learning processing platform. A new software-hardware co-design workflow including model compression, compilation, and hardware acceleration is proposed. Two architectures named Aristotle and Descartes are designed to accelerate CNN and RNN respectively. Results show that, considering the optimization in both software and hardware part, it is possible to achieve over 10x energy efficiency compared with GPU even on FPGA. The experiences in the journey from technologies to a real product and starting a business will also be shared.

About Song Yao:

Song Yao is the CEO and Co-Founder of DeePhi Tech, a startup that is devoted to provide the world with more efficient deep learning platform. He is a well-recognized researcher in hardware acceleration of deep learning. Before founding DeePhi Tech, he was a visiting researcher in Stanford University and received his B.S. Degree in Tsinghua University in 2015. He has received many awards including FPGA 2017 Best Paper, Top 30 AI Entrepreneurs in China, and Forbes 30 Under 30 Asia.

Hosted by: Yuan Xie