The Seung Lab

Ila Fiete

Graduate Student
(617) 452-2691
prasad@physics.harvard.edu

Oculomotor Integrators and Persistent Activity

The generation of persistent neural activity in response to a transient stimulus is implicated in a host of brain functions, for example, the ability to command eye muscles to stay contracted (and hence keep the eye fixed at some eccentric position), the capacity to retain a short-term memory of the location of objects viewed briefly in a subsequently darkened room, and the ability, over short times, to maintain a sense of direction while wandering around in a featureless landscape.

The time-scales involved in such activities are on the order of 10 seconds; however, microscopic persistence times of typical, isolated neurons in vertebrate CNS are 100 times smaller, of order 100ms. Some models of networks with recurrent excitation successfully replicate critical features of persistent activity and more generally, carry out integration of transient inputs signal, also a very important feature, for example, in the control of eye movement and the compensation of head velocity in vision. However, endemic to these recurrent feedback models is a requirement for stringent tuning of the parameters. To obtain experimentally observed persistence times, the parameters of the model must be carefully adjusted to lie within 1% of the tuned values.

I am looking for plausible (i.e., biologically relevent) schemes to improve the generic robustness of this class of models against small perturbations of the parameters away from their tuned values. In addition, I am interested in and working on the problem of learning, as applied to these (oculomotor) integration areas in the brain.