NSF awards Illinois funding for a deep learning research instrument

The rapid development of deep neural networks and their function in artificial intelligence, machine learning, and computer vision is transforming many disciplines. With applications in voice and image recognition, speech processing, healthcare, education and a plethora of other arenas, teaching computers to behave like the human brain, more than simply compute, is essential to technological advancement.

You may not think about it when you speak to your phone's virtual assistant, tag a picture on Facebook or use an app to remove your background in a video-chat, but all of those tasks use some form of "deep learning," where computers are taught to recognize situations and act upon them, inspired by how living brains work. In order to be able to teach computers to think rather than compute, however, a system that allows researchers to explore ways to harness multiple systems in developing and applying deep learning algorithms is essential.

In order to lay the groundwork for these deep neural networks, and in turn study deep learning at a large scale, CS @ ILLINOIS Professors William "Bill" Gropp, Roy H. Campbell, and Jian Peng, with Volodymyr Kindratenko, of the National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign, have been awarded over $2.7 million from the National Science Foundation (NSF) Major Research Instrumentation (MRI) program to build a dedicated research instrument to expand deep learning research.

"This deep learning instrument will bolster current relevant deep learning research communities here at the University of Illinois, allowing researchers to leverage deep learning more than they ever could before," said Bill Gropp, who is also the director of the NCSA and holds the Thomas M. Siebel Chair in Computer Science. "This NSF grant will allow the University of Illinois to expand deep learning research opportunities to a wider group of interested researchers, including undergraduates, and will help to develop the deep learning workforce."

The deep-learning instrument will open the door to new industry-academia collaborations that have never before been possible. Furthermore, the blueprint of the system architecture and the instrument will be publicly available, allowing varied domains to build off of these deep learning frameworks.

"The NSF-funded deep learning research instrument grant will support our excellent university research team as they investigate new systems and algorithms for machine learning and data analytics that extract actionable knowledge from the massive sets of data available today and will contribute to new scientific and commercial applications," said Campbell, a co-principal investigator.

"The scalability, sharing of resources, and ease of use of these tools is essential to encourage widespread deployment in the industrial environment," continued Campbell, the Sohaib and Sara Abbasi Professor of Computer Science. "As an example of some of the applications that the grant will enable, the new system will accelerate our work in analyzing neurological disorder multi-modal datasets including genomic, imaging, and clinical information. The system will enable us to analyze these complex datasets and develop new discoveries and predictive models."

The new deep learning infrastructure will be constructed in collaboration with IBM and NVIDIA.