Research
Computational
Neuroscience and Brain-Computer Interfaces
The primary goal of my research is to discover the
computational principles underlying the brain's remarkable ability to learn,
process and store information, and to apply this knowledge to the task of
building adaptive robotic systems and brain-computer interfaces (BCIs). How
does the brain learn efficient representations of objects and events occurring
in the natural environment? What are the algorithms that allow useful
sensorimotor behaviors to be learned? What computational mechanisms allow the
brain to adapt to changing circumstances and remain fault-tolerant and robust?
How can the knowledge gained through computational studies of the brain be used
in biomedical applications such as BCIs for the disabled? My students and I are
investigating these questions using a combination of probabilistic techniques,
computer simulations, and collaborative neurobiological experiments. Such an
interdisciplinary approach has provided functional interpretations of several
otherwise puzzling neurobiological properties while at the same time suggesting
biologically-inspired solutions to problems in computer vision, robotics and
artificial intelligence.
Humanoid Robots
that Learn from Humans
We are developing new methods that allow a humanoid
robot to learn new actions and skills from a human teacher in much the same way
that human infants and adults learn through observation and experimentation.
Such an approach opens the door to a potentially powerful way of programming
general-purpose humanoid robots--through human demonstration, obviating the
need for complex physics-based models and explicit programming of behaviors.
Analysis of the
4000-year-old Indus Script
Despite a large number of attempts, the script of the Indus civilization (circa
2500-1900 BC) remains undeciphered. The absence of a
multilingual "Rosetta stone" as well as our lack of knowledge of the
underlying language have stymied decipherment efforts. Rather than attempting
to ascribe meaning to the inscriptions, we are applying statistical techniques
from the fields of machine learning, information theory, and computational
linguistics to first gain an understanding of the sequential structure of the script.
The goal is to discover the grammatical rules that govern the sequencing of
signs in the script, with the hope that such rules will aid future decipherment
efforts.