[Abstract]

In the past few years, wide applications of deep neural networks (DNN) have contributed to significant progress in various tasks such as image classification, object detection, and segmentation. Particularly, DNNs are inevitable SW stack in autonomous vehicles on the high performance computing platform such as GPU. In order to achieve higher accuracy and lower infernece, DNN optimization process is required from DNN design to training, and to deployment. In this talk, several DNN optimization approaches and techniques will be presented: sparsity, neural architecture search, DNN quantization for lower precision inference, workflow of model conversion, GPU scheduling overhead, etc

[Biography]
Dr. Su Inn (Josh) Park is currently a Automotive Devtech / Solutions Architect Manager at Nvidia Corporation. He received B.S and M.S. degrees from Korea University, Seoul, Korea, and Ph.D. degree from Texas A&M University, College Station, TX in Computer Science. He has successfully accomplished various projects with varied academic backgrounds and professional industrial experience.  Before his Ph.D., he was a software engineer at Samsung Electronics, and during his Ph.D., he was a research intern at LG Electronics, Samsung Austin R&D Center, NCTR  FDA, and Nvidia. Before joining Nvidia, he was a data scientist at Samsung Austin Semiconductor analyzing big data on Apache Hadoop and Spark. To date, he has been working on deep learning solutions using open-source framework such as TensorFlow on muti-GPUs/multi-nodes servers and embedded systems. Also, he has been evaluating and improving training and inference performances on various GPUs + x86_64/aarch64.

ZOOM : https://zoom.us/j/8978217407?pwd=d1pOZmF1OWlseEdZRVBpV3VuSkl3dz09

ID : 897 821 7407

PW : 1nTQDY