Winner-Take-All Network
The neuron with the highest weighted input
in a winner-take-all network takes on an upper value, all other neurons
take on the lower value. In case of a tie the node that wins is chosen
randomly.
Initialization
When creating a WTA network you can specify how many neurons are in the network and also values for the parameters listed below. The standard layouts are also avaialble.
Parameters
Number of Units: The number of units in the network.
Lower Value: The activation of any node which does not have the greatest weighted input.
Upper Value: The activation of the node which has the greatest weighted input.
Set Winner Randomly: If selected, there is some chance that the winner is chosen randomly.
Probablity of choosing a random winner: When set winner randomly is selected, the is the probabilyt of randomly choosing a winner.