Minsoo Rhu is an assistant professor at the School of Electrical Engineering at KAIST, South Korea. He was formerly with NVIDIA Research as a senior research scientist developing hardware & software systems for accelerating deep learning algorithms. He holds a B.E. from Sogang University (2007), a M.S. from KAIST (2009) and a Ph.D. from the University of Texas at Austin (2014).
강연요약 :
Deep learning (DL) is currently the fastest-growing field in machine learning and is transforming various segments of our lives. Today, deep learning applications are most commonly trained using GPUs and are then deployed for inference on top of a wide range of systems from mobile devices to datacenters. This talk will first discuss the basic design principles behind the GPU hardware and software architecture. I will then explain why GPUs have become the state-of-the-art platform for accelerating DL training, followed by a brief discussion on some of the key challenges of training/deploying emerging deep learning applications.
