UZH ETH UZH|ETH
  Zurich

Skip to content. | Skip to navigation

Sections
Personal tools
You are here: Home Projects neuroP Hardwired Coupled WTA

Hardwired Coupled WTA

by Lorenz Muller — last modified Nov 28, 2012 12:28 AM

We investigated a chip architecture of coupled hardwired WTA, in which basic small WTA units can be brought into cooperation or competition by a series of hierarchical switches. For this layout we followed Rutishauser, Slotine and Douglas 2012. 

hardwired_connectivity

Figure 1: 

Top: Schematic wiring diagram of the neurons, closing the black switch increases population size, the red switch puts population into competition

Bottom: (Left) Population level diagram, (Right) Hierarchical switches.

cooperate_compete_WTA 

Figure 2: Simulation

Top: Rasterplot; one of the 4 competing modules, each consisting of 4 WTA units, wins the competition and assumes a self-sustained activity state.

Bottom: (Left) Initial input, provided for 150 ms, (Right) The population activities of the units.

The advantage of joining multiple units into a cooperating population, is that in this configuration the variability in their parameters can even out over the population. However this effect is only strong for small population sizes. 

Computationally coupled WTA are interesting (among other reasons) because they can easily be configured to have ghost attractors from which arbitrarily small probabilistic perturbations lead to different attractor states.

ghost_attractor

Figure 3: Flow-Diagram of two coupled WTA

At ca. (20,20) there is a ghost attractor for the region x,y>20