...It does everything a conventional neural network does,
just at a higher level of abstraction.
In a conventional network, information enters at the synapses and neurons fire or
not. With a Fractal Neuro Oscillator, the synapse connections are randomized and
information enters by firing neurons and then measuring the percentage of time
firing of select neurons to generate an output value. These oscillation states can
be summarized into a digest pattern to represent a saved state. Gradient error values can then be calculated.
A Fractal Neuro Oscillator also can be trained to oscillate in a repeating manner
via. a random training algorithm.
...