Home
History
INS News
Weekly Events
Seminars
Community Programs
The Stellar Program
INS Members
Neuroscience Graduate Group
Other Educational Activities
Society for Neuroscience
INS Members
Kwabena
A. Boahen, Ph.D.
Asst Professor,
Depts of Bioengineering
and Electrical Engineering
School of Engineering and Applied Science
120 Hayden Hall/6392
(215) 573-4072 FAX: (215) 573-2071
email: kwabena@neuroengineering.upenn.edu
More information
about Dr. Boahen
Click here for selected publications since Dr. Boahen's arrival at Penn
RESEARCH INTERESTS
Very large-scale models of entire neural systems consisting of thousands
of silicon neurons that respond in real-time. RESEARCH TECHNIQUES
Very large-scale models of entire neural systems consisting of thousands of silicon neurons that respond in real-time.RESEARCH SUMMARY
How come the brain uses a million times less energy per operation than computers do? No one knows for sure. Is it because the brain is analog whereas computers are digital? Not quite: The brain uses digital axonal spikes as well as analog dendritic potentials.My research suggests that it is because the brain is 'programmable' at the level of individual connections-it learns by growing. By morphing such customized neural circuits into a silicon chip, my lab developed a silicon retina that computes a thousand times more energy-efficiently than a computer does. Although this chip achieves desktop-computer performance while consuming palmtop-computer power, it is still three orders of magnitude shy of the retina's energy-efficiency. But this gap is closing with the rapid miniaturization of silicon technology, which will soon enable us to put a billion transistors on a single microchip.
A billion-transistor chip presents a daunting 'programming' task similar to that nervous systems had to solve. It takes ten quadrillion (1016) bits to specify how a quadrillion (1015) synapses connect the brain's one trillion (1012) neurons-but the entire genome contains only billion (109) bits. Having exhausted all information available in the genetic code, neural circuits customize themselves further through interaction with their internal and external environments. By mimicking such epigenetic development, we are building a 'computer' that programs itself at the level of individual connections. In particular, my lab is currently morphing the retinotectal, geniculocortical, and hippocampal systems into silicon by mimicking activity-dependent, neurotrophin-guided, anatomical plasticity, as well as the biophysical differentiations of various neuronal types.
KEY WORDS:
Neuroengineering; Neuromorphic; retinomorphic; computational neuroscience; neural computation; spike coding; neural networks; vision; audition
