NeuroThursday Spring 2003

NeuroThursday Spring 2003

10:30 AM - 12:00 Noon Thurdays in Warren Weaver Hall Room 1314

February 13: Jack Schwartz (NYU) Some psychophysical experiments on the perception of Glass patterns visual motion, and the Cafe Wall illusion, with some observations on the relationship between psychophysics and direct brain probing

February 27: Ken Miller (UCSF) TBA

March 13: Souheil Inati (NYU) The inverse problem formulation of MRI as a large scale optimization with some example applications from neuroscience

April 3: Carson Chow (Pittsburgh) A spiking neuron model of binocular rivalry

April 10: Valeria Del Prete (Kings College London) A theoretical model for population coding of mixed continuous and discrete stimuli: from the linear rise to the asymptotic regime

April 24: Christopher Pack (Harvard Medical School) TBA

The Computational Neuroscience Forum & CIMS NeuroThursday Joint WebPage


Valeria Del Prete(Kings College London) A theoretical model for population coding of mixed continuous and discrete stimuli: from the linear rise to the asymptotic regime

In real experiments animals are exposed to a discrete set of stimuli or trained to perform a discrete set of tasks. Yet, natural stimuli are multi-dimensional and can be parameterized by a mixture of continuous and discrete dimensions. For example an arm movement can be categorized according to its (discrete) "type"- e.g. bimanual or unimanual- and its (continuous) direction. Inspired by data recorded by E. Vaadia from the SMA and M1 areas of monkeys performing such arm movements, I have developed a theoretical model of a feed-forward network of threshold-linear neurons to investigate the information transmitted between the two areas in presence of continuous+discrete input patterns. I have studied analytically the information carried by the population in the first layer in the network with respect to the complex stimuli, showing that in the regime of the initial information rise as a function of the population size, there is no difference in using a more realistic model of the firing or a simple gaussian model with respect to the fit of real data: the analytical expression is the same except for a renormalization of the noise. On the other hand an analytical evaluation of the information in the asymptotic regime reveals a more complicated relationship between the two models. Introduction of correlations in the neurons' preferred orientations (correlations actually observed in the data) does improve the fit suggesting that the correlations detected in the data are information bearing. Then I have evaluated the information carried by the output layer in the limit of a large number of input units, showing that, in absence of learning in the connectivities, the presence of the output noise and of a threshold are two simple mechanisms which cause an information loss in output.

Carson Chow (Pittsburgh) A spiking neuron model of binocular rivalry

Binocular rivalry occurs when the two eyes are presented with drastically different images. Only one of the images is perceived at a given time, and every few seconds there is alternation between the perceived images. The perceived durations of the images are stochastic and uncorrelated with previous perceived durations. Recordings from single neurons in the visual cortex of monkeys experiencing binocular rivalry find that neuronal activity is correlated with the monkey's perception. I will present a biologically plausible network of Hodgkin-Huxley type neurons that exhibits rivalry and can account for the experimental and psychophysical results. A reduced population rate model can be derived from the spiking model from which a simple explanation of the model dynamics can be obtained.

A Recap of NeuroMonday Fall 2002

Louis Tao, My CIMS web page, Applied Mathematics Laboratory & CNS, New York, NY 10012. This page last updated March 2003.