Home
Contact
Short CV
Research
Publications
Software
Links
Impressum

University of Plymouth | Centre for Theoret. and Comput. Neuroscience | Home   

Example 6: Simple V1 Model

This page presents a simple model of the primary visual cortex. It does not aim at being very sophisticated, but only at showing what is possible using Felix.

The figure below displays a snapshot of the running simulation program.

The model consists of two square layers (a 63*63 neurons, ca 8000 cells in total) of probabilistically spiking neurons or, alternatively, graded response neurons. The type of neurons can be selected while the simulation is running by pressing the respective button ("spikes") in the top control panel shown on the right.

The two fields comprise excitatory and inhibitory neurons, respectively. They are coupled locally within certain neighbourhoods by Gaussian connectivity kernels (E to E, E to I, and I to E) -- the sliders J11, J21, and J12 control the respective coupling strengths. In addition, the cells are connected by a sparse network of connections with distance dependent propagation delays (not seen in the figure) -- the slider Jlr sets the respective weight-level of long-range synaptic couplings.

The excitatory cells in the model are orientation tuned, ie, each cells is equipped with a Gabor-type receptive field. The OT map and one of the RFs is shown at the top of the above figure. The sparse long-range connections are predominantely between cells with similar orientation tuning.

Input to the network model is either a disc of variable size and contrast (not shown; can be switched on by one of the top-row buttons in the control panel), or a moving or counterphase grating. Parameters of the disc/grating can be controlled by sliders in the control panel while the simulation is running (orientation, size, wave-number, temporal frequency).

The four large canvases show the input stimulus grating, Gabor filtered input signals, potentials, and output spikes at simulation time step 1892. Observe, how the filtered inputs and potentials reflect the OT map (for the given stimulus orientation). If stimulus orientation is changed, the activated regions move accordingly. If the size of the stimulus is made smaller, only activity in the central region remains, but that in the outer areas goes to zero. 1000 simulation steps of the model need about 30s and are roughly equivalent to 1s real-time.

Downloads

Statically linked executable (Right-click and store using "Save Link Target .." or similar. Start by clicking on it after download.)

Orientation tuning map necessary to run the example. (Right-click and store using "Save Link Target .." or similar to the same directory as the executable. )

Environment file (Put this file into a subdirectory "env" in the directory of the executable, which will then start with appropriate parameters.)

   © 2007 by -thowe-