Title: Building Systems Technology for Deep Learning
Speaker: Dr. Sanket Tavarageri (Intel Labs)
Host Faculty: Dr. Ramakrishna Upadrasta
Room No: C-LH5
Time: 11:00 - 12:00
Machine Learning, and especially Deep Learning has revolutionized computing industry by automating tasks which were previously strictly in the purview of humans: for example, deep learning can perform image recognition, language translation, autonomous driving, and speech recognition with great accuracy. The Deep Neural Networks (DNNs) which lie at the heart of deep learning learn patterns from data. The use of DNNs is compute and data intensive which is prompting innovations in the entire hardware and software stack to make their application practical and use them for ever more complex tasks.
In this talk, I will introduce the basic concepts of DNNs, and describe our ongoing work in building an effective compiler specifically for deep learning. We will also overview the other prominent algorithmic, and architectural innovations taking place in industry, and academia targeted for deep learning.
Dr. Sanket Tavarageri is a Research Scientist at Intel. He obtained a Ph.D from The Ohio State University for his work developing compiler technology for achieving high performance, and fault tolerance in applications. His current work includes developing optimizations for deep learning in various frameworks targeting bleeding-edge Intel architectures.
Dates: Tuesday, February 12, 2019 - 11:00 to 12:00