Solution Topology
- One of the benefits of setting the neural parameters is an improved
methodology.
- There are fewer free parameters.
- Fatiguing to spontaneous firing also simplifies the method. Now
we don't have to arbitrarily fire a neuron.
- Saturation base (the desired weight for input to a neuron) is 4
in all nets; learning rate is .01.
- However, there remain a lot of free parameters (practically
infinite), with the largest freedom in topology.
- The topology used was a three layered system
- Input is 100% excitatory neurons.
- Gas and Output is 50% excitatory 50% inhibitory.
- Each neuron has 20 connections inside the subnet randomly selected.
- There were connections from the input to the Gas.
- There were connections from the Gas to and from the Output.
- Each excitatory neuron has 10 connections randomly selected
to the other net.
