| We propose a new theoretical model of Artificial Neural Networks (ANN'S) based on well established ideas in the field of Cellular Automata (CA), the field of Quantum Mechanics (QM) in general, and the field of Quantum Electrodynamic (QED) in particular. The model uses a new architecture where neurons (nodes or units) are located within a 3-D cubic lattice. The connections between the nodes are well defined. The input to the model is a set of patterns encoded as waves packets propagating across the whole set of nodes. The propagation of those waves is analogous to the process of waves scattering across a set of potentials (here potentials exist at the level of each node). Storage and retrieval in this model is based on 3-D scattering. A node behaves as a quantum harmonic oscillator, a feature by which a propagating wave would leave a piece of information it encodes as a set of oscillations at the node level. |
