Molecular-junction-nanowire-crossbar-based neural network
View Patent ↗A method for configuring nanoscale neural network circuits using molecular-junction-nanowire crossbars, and nanoscale neural networks produced by this method. Summing of weighted inputs within a neural-network node is implemented using variable-resistance resistors selectively configured at molecular-junction-nanowire-crossbar junctions. Thresholding functions for neural network nodes are implemented using pFET and nFET components selectively configured at molecular-junction-nanowire-crossbar junctions to provide an inverter. The output of one level of neural network nodes is directed, through selectively configured connections, to the resistor elements of a second level of neural network nodes via circuits created in the molecular-junction-nanowire crossbar. An arbitrary number of inputs, outputs, neural network node levels, nodes, weighting functions, and thresholding functions for any desired neural network are readily obtained by the methods of the present invention.
1. A nanoscale neural network circuit comprising:
input nanowire signal lines;
output nanowire signal lines; and
at least one layer of neural network nodes programmed into a molecular-junction-nanowire crossbar.
2. The nanoscale neural network circuit of claim 1 wherein the neural network nodes are programmable.
3. The nanoscale neural network circuit of claim 1 implemented in a complementary/symmetry lattice additionally including other nanoscale components that, together with the nanoscale neural network circuit, compose an electrical subsystem.
4. The electrical subsystem of claim 3 wherein the nanoscale components are configured together at densities within the electrical subsystem at densities greater than 1.0 giga-transistors/cm 2 .
5. The nanoscale neural network circuit of claim 1 wherein each neural network node comprises:
a set of resistors programmed into the molecular-junction-nanowire crossbar to weight selected input signals and combine the weighted input signals onto a nanowire carrying the weighted, combined signal S; and
a thresholder programmed into the molecular-junction-nanowire crossbar comprising a selectively configured pFET and a selectively configured nFET that thresholds the signal S.
6. The nanoscale neural network circuit of claim 1 wherein the resistances of the resistors is chosen to provide a precise weighting of the input signals combined to produce signal S.
7. The nanoscale neural network circuit of claim 1 wherein microscale non-semiconductive signal lines are used in place of nanoscale non-semiconductive signal lines.
8. A method for configuring a nanoscale neural network circuit, the method comprising:
providing a molecular-junction-nanowire crossbar with microregions prepared for selective configuration of pFETs, nFETs, and connection elements;
selecting a number of input nanowire signal lines;
selecting a number of output nanowire signal lines; and
programming electrical components at the junctions of a molecular-junction-nanowire crossbar to construct at least one layer of neural network nodes.
9. The method of claim 8 further including:
configuring each neural network node by:
programming a set of resistors to weight the selected input signals and to combine the weighted input signals onto a nanowire carrying the weighted, combined signal S; and
programming a pFET and an nFET that inverts the signal S.
10. The method of claim 9 wherein selectively configuring a pFET and an nFET that inverts the signal S further includes:
selecting resistances of the selectively configured resistors to provide a precise weighting of the input signals combined to produce signal S.
11. A nanoscale neural network circuit comprising:
input nanowire signal lines;
output nanowire signal lines; and
at least one layer of neural-network-node means.
12. A nanoscale thresholding circuit comprising:
an input nanowire signal line;
an output nanowire signal line; and
junction components programmed into a molecular-junction-nanowire crossbar.
13. The nanoscale thresholding circuit of claim 12 wherein the junction components are programmable.
14. The nanoscale thresholding circuit of claim 12 implemented in a complementary/symmetry lattice additionally including other nanoscale components that, together with the nanoscale thresholding circuit, compose an electrical subsystem.
15. The electrical subsystem of claim 14 wherein the nanoscale components are configured together at densities within the electrical subsystem at densities greater than 1.0 giga-transistors/cm 2 .
16. The nanoscale thresholding circuit of claim 12 wherein the junction components include:
an nFET transistor connecting the input nanowire signal line to a relatively high voltage source; and
a pFET transistor connecting the input nanowire signal line to ground.
17. The nanoscale thresholding circuit of claim 16 wherein the junction components further include:
a first interconnection interconnecting the output nanowire signal line to a relatively high voltage source; and
a second interconnection interconnecting the output nanowire signal line to ground.