Prior to his recent appointment as a postdoctoral fellow at SFI, Artemy Kolchinsky worked with EEG and functional MRI data, looking at how highly integrated the brain is on different scales. He says he has always been fascinated by the relationship between complexity and cognition.
“The brain is the archetypal complex system,” says Kolchinsky, who hopes to use novel mathematical techniques to understand ever-increasing amounts of brain imaging data.
While at SFI, Kolchinsky is working with SFI Professor David Wolpert on several projects related to optimal use of information and prediction. One is the problem of modeling and analyzing complicated dynamical systems that require large amounts of time and computational power to simulate. An example would be the propagation of disturbances and blackouts on an electrical grid. Their question: Given such a system, how can we find a compression of it that still gives us good predictions but is much cheaper to run? Another project investigates connections between information processing and statistical physics. A longstanding notion is that to perform a computation, a minimum amount of energy is required.
“David demonstrated that the amount of entropy dissipated can depend not only on the function, but what you expect the inputs to be,” says Kolchinsky. “There is a certain thermodynamic cost to making the wrong predictions.”
The researchers are working toward generalizing and extending these results. Finally, the two are beginning to work on understanding why different social groups develop different organizations, whether the group is a prehistoric tribe or a business firm. Here again appears the idea that information processing is a costly resource. Some types of organizations appear to use this resource to coordinate group activity more efficiently than others.
Prior to joining SFI, Kolchinsky received a PhD in informatics (with focuses on complex systems and cognitive science) from Indiana University.