By Nicole Contosta
Obtaining knowledge, growth, and breaking boundaries lie at the forefront of NSF recipient Keshava Katti’s approach to research with biological neuron model.
Specifically interested in spiking neurons, Keshava’s work rests in developing a deeper understanding of the brain, especially regarding how neurons achieve different tasks, as well as how such ensemble behavior can yield design principles for machine learning algorithms. “What we call a neuron in an artificial neural network is very crude, it can barely be considered a neuron. Most people [in the field] laugh off serious comparisons to the brain,” Keshava said.
So how does this goal translate into tangible objectives? By using biological neuron models in learning algorithms, Keshava studies how to deduce information from spatiotemporal data. He further considers how these algorithms could give rise to a new class of neuromorphic or synaptic-inspired devices for computation. One such class is composed of retinomorphic devices, which, when implanted in a robot, could yield artificial vision capacity that senses changes in light intensity and performs basic mathematical operations in situ. “The devices themselves would be like a camera, memory, and computational unit in one, which could in turn help navigate an unknown environment just as our retina and visual cortex would,” Keshava said.
Keshava’s research takes an interdisciplinary approach. He is co-advised by Dr. Deep Jariwala in Nanodevices and Nanosystems and Dr. Pratik Chaudari in Information and Design Systems. According to Keshava, Professor Chaudari has the theory of deep learning and an avid interest in computational neuroscience, while Professor Jariwala has the expertise to design specialized hardware.
Fusing different fields of engineering and science has roots in Keshava’s childhood. His father, Rajendra, was a professor of electrical engineering at North Dakota State University. His mother, Nandini, was a seventh-grade life science teacher. He remembers visiting his father’s office and staring at the board’s mysterious equations with fascination, wanting to understand how it all worked. In his mother’s classroom, he recalls helping her prepare for an activity on human evolution. This piqued his curiosity. “Humans know so much about their evolution and cannot yet harness this to create truly intelligent machines,” Keshava said.
In his free time, Keshava enjoys DJing and in fact formed the organization DJ@Penn, open to all Penn students. There, he leads training on how to book gigs at clubs as well as how to utilize DJing software. Recently, he purchased a turntable. Once he masters the craft, he wants to teach others the old-school art of spinning records.
Education, in all its forms, remains a lifelong pursuit for Keshava. One must look no further than one of his original sources of inspiration— his mother —for confirmation of this fact. At 60, Nandini earned her Doctorate in Pharmacy.