Training of a physical neural network

A physical neural network system based on nanotechnology, comprising:

• an electromechanical-based physical neural network configured utilizing nanotechnology, wherein said electromechanical-based physical neural network comprises at least one synaptic component comprising a plurality of nanoconductors suspended and free to move about in a dielectric solution comprising a mixture of a dielectric solvent and said plurality of nanoconductors, wherein said plurality of nanoconductors comprises neural network connections in said dielectric solution between at least one pre-synaptic electrode and at least one post-synaptic electrode that provide said at least one synaptic component of said electromechanical-based physical neural network;
• a mechanism for dynamically implementing variable connections from said electromechanical-based neural network connections, wherein said mechanism comprises a dielectrophoretic assembly of said plurality of nanoconductors in said dielectric solution, wherein said plurality of nanoconductors suspended in said dielectric solution are subject to a dielectrophoretic force resulting from an exposure to a time-varying electric field, such that said dielectrophoretic force includes a dipole-induced force that is utilized to attract or repel said plurality of nanoconductors to an electrode gap between said at least one pre-synaptic electrode and said at least one post-synaptic electrodes;
• and
• a feedback circuit for training said electromechanical-based physical neural network by providing at least one feedback signal to said physical neural network for modifying said at least one synaptic component of said electromechanical-based physical neural network to accomplish a particular neural network task based on a non-supervised Anti-Hebbian and Hebbian (AHAH) rule, thereby permitting said electromechanical-based physical neural network to provide for a physical process capable of self-assembly and self-repair and wherein said electromechanical-based physical neural network functions as a physical mechanism that emulates said non-supervised AHAH rule.

A physical neural network method based on nanotechnology, comprising:

• configuring a electromechanical-based physical neural network having at least one synaptic component, wherein said at least one synaptic component of said electromechanical-based physical neural network comprises a dielectric medium with a plurality of nanoconductors disposed and free to move about in said dielectric medium, said dielectric medium comprising a dielectric solution comprising a mixture of a dielectric solvent and said plurality of nanoconductors;
• modifying a strength of said at least one synaptic component by subjecting, said plurality of nanoconductors s suspended in said dielectric medium to a dielectrophoretic force resulting from an exposure to a time—varying electric field, wherein said dielectrophoretic force includes a dipole—induced force that is utilized to attract or repel said plurality of nanoconductors to an electrode gap formed between at least one pre-synaptic electrode and at least one post-synaptic electrode that form said at least one synaptic component of said electromechanical-based physical neural network;
• and
• training said electromechanical-based physical neural network utilizing a feedback circuit, which provides a feedback signal to said electromechanical-based physical neural network, thereby by modifying said at least one synaptic component of said electromechanical-based physical neural network to accomplish a particular neural network task based on a neural network training rule, thereby permitting said electromechanical-based physical neural network to provide for a physical process capable of self-assembly and self-repair and wherein said electromechanical-based physical neural network functions as a physical mechanism that emulates said neural network training rule.