This was presented at the CNS2000 computational neuroscience meeting,
in Brugge, Belgium, July 15-20, 2000.
Stochastic optimal control and the human oculomotor system
Liam Paninski, Michael J. Hawken
Nervous systems seem to efficiently perform a large class of computationally difficult statistical tasks. We rigorously quantified the efficiency of the oculomotor system's solution to a prototypical example of this type of problem, with the hope of deriving a deeper understanding of how our brains solve such problems in general. Specifically, we observed the eye movements of human subjects attempting to track a visual target that moved stochastically across a computer screen. The subjects were able to perform surprisingly near the optimum under the conditions examined - an important step in verifying the introspective conviction that nervous systems are somehow designed to compute physically relevant probabilistic objects.
Here is a manuscript (ps.gz, 144 K), 16 postscript pages + 7-8 (color) figures, summarizing the work to date.
A shorter proceedings paper was published as "Stochastic optimal control and the human oculomotor system," Paninski L and Hawken M, Neurocomputing 38-40: 1511-1517 (2001).
Liam Paninski's home