GATSBY COMPUTATIONAL NEUROSCIENCE UNIT
UCL Logo

Klaus Obermeyer

Department of Electrical Engineering and Computer Science, Berlin University of Technology, Germany

 

Wednesday 22 March 2006

16:00

 

Seminar Room B10 (Basement)

Alexandra House, 17 Queen Square, London, WC1N 3AR

 

 

Computational Approaches to Adaptation

 

In my talk I will first present modelling results from a follow-up study of [1]. Using a firing rate and HH-neuron network models we analysed (1) how much evidence the recent intracellular measurements from cat primary visual cortex provide for different cortical operating regimes and (2) whether the available data allows to single out a cortical "operating point". Using a Bayesian analysis we find, that the experimental data most strongly support a regime where the afferent input is well tuned but where the local cortical network provides dominant excitatory and inhibitory recurrent inputs (compared to the feedforward drive). Interestingly, neither Mexican-hat type interactions nor a particular spatial extend of the cortical excitatory vs. inhibitory connections have to be invoked to draw this conclusion. We also find, that the most likely operating point is close to the border with and unstable regime or with the so-called marginal phase. Hence it is conceivable, that modulatory effects may briefly shift the operating point into these regimes.

 

Secondly, I will present results of a computational study of adaptation in a recurrently connected visual cortical hypercolumn. Using Fisher information as a criterion for optimality I will show, how changes in tuning curves during adaptation depend on architectural constraints, for example on the postulated site of plasticity. Results will then be discussed in the context of the - quite different - phenomena perceptual learning and attention.

 

Finally, I will report results of a computational study on a context dependent mapping of visual information ("representation") into action ("readout"). Motivated by experiments like the Stroop task I will illustrate, how task-dependent feedback can select the relevant stimulus dimensions for a task at hand, which then dominate the responses of neurons initiating the actions.

 

[1] Marino, Schummers, Lyon, Schwabe, Beck, Wiesing, Obermayer & Sur, Nature Neurosci. 8, 194ff [2005].