Illinois researchers studying machine learning and perception as part of new Intel-funded center

Three University of Illinois engineering faculty members will be investigating machine learning and perception via stochastic computation as part of a new Intel-funded center on embedded computing. Computer Science Professor Rob Rutenbar, ECE Professor Naresh R Shanbhag, and ECE Assistant Professor Paris Smaragdis (who has a joint appointment with Computer Science) are part of the Intel Science and Technology Center on Embedded Computing headquartered at Carnegie Mellon University.

With the growing popularity of mobile real-time and personalized technology, there is a corresponding rise in demand for specialized embedded computing systems to support a broad range of new applications— including many not yet envisioned.

A key area of research for the center is to make it easier for these everyday devices to continuously collect, analyze and act on useful data from both sensors and online databases in a way that is timely, scalable and reliable. For example, in cars, this data could be used to customize in-vehicle entertainment options when specific passengers are recognized, and provide them better routing, retail, dining, and entertainment recommendations while on the road.

Tapping into the expertise of leading researchers from Carnegie Mellon University, Cornell University, University of Illinois at Urbana Champaign, University of Pennsylvania, Pennsylvania State University, Georgia Institute of Technology, the University of California at Berkeley, and Intel, the ISTC for embedded computing forms a new collaborative community to drive research to transform experiences in the home, car and retail environment of the future.

As part of the overall effort, the Illinois team will integrate three broad themes of research work: silicon-based machine learning accelerators (Rutenbar), stochastic silicon (Shanbhag), and machine learning and perception (Smaragdis).

The team’s research will be informed by new applications of machine listening in embedded computing environments for cars, retail environments, and homes. For example, the Illinois team’s efforts may lead to cars that can listen for driver activity such as sleepiness, mechanical issues, or external traffic cues such as accidents. In a retail environment, the Illinois team envisions applications that can listen for customer movement and patterns linked to in-store advertising or product placement, or for emergency situations. Home applications may include listening for accidents and emergencies for the elderly or those needing assistance, or for building systems malfunctions like broken pipes.

“These new ISTCs are expected to open amazing possibilities,” said Justin Rattner, Intel Chief Technology Officer. “Imagine, for example, future cars equipped with embedded sensors and microprocessors to constantly collect and analyze traffic and weather data. That information could be shared and analyzed in the cloud so that drivers could be provided with suggestions for quicker and safer routes.”