Overview | HNeT2005 | Performance Aspects | Mathematics | Biology
 

The core learning method applied within HNeT is shown to the right.  Information within these matrices may be represented by frequencies, quantum states, spike interval/amplitude, etc… in all practicality they are sets of complex numbers or complex manifolds.  By applying the HNeT learning process, a trained neuro-holographic cell will generate a deterministic response to each prior learned memory trace as shown below.


An interesting property occurs in the event that a cell or assembly containing N cortical memory elements is trained beyond the point of saturation.  This being that mean magnitude in memory [X] approaches an asymptote of sqrt(N), while memory traces continue to accumulate unattenuated.  This asymptotic behavior is shown to the left.

The implication being that complex manifolds of finite resolution can retain an unbounded quantity of associative memory traces.