Comments
Loading Dream Comments...
You must be logged in to write a comment - Log In
ArtistTwo lines of carbon atoms. Each atom in the first line is bonded to every atom in the second line. The bond is shown as a dotted line.
First, re-read the description in the original drawing (MSM's versus NN's) because browser limitations prevent its copyal to here. Now, consider the first row. A branch connects each output to every input in the second row. In a NN these branches have the weighting factors, which are adjusted by training. In a MSM these branches carry the probability of a state transition (or a dataflow stimulus) to each of the nodes in the second row, again established by training. If we provide sufficient similar stimuli to the MSM and appl y superposition to the outputs (responses) we essentially have a fair simulation of a NN. We can do this by running several parallel threads through the states of the MSM, just like the famous double-slit experiment fires photons though multiple paths. We may even be able to construct a quantum AI engine in this manner, but that's an area for further research. E&OE. :)