Wireless neural network and a wireless neural processing element

• US 5,297,232 A
• Filed: 10/30/1991
• Issued: 03/22/1994
• Est. Priority Date: 10/30/1991
• Status: Expired due to Fees

• Current Assignee: Northrop Grumman Corporation
• Original Assignee: Westinghouse Electric Company LLC (Brookfield Asset Management Incorporated)
• Inventors: Murphy, John H.
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/785,393
• Time in Patent Office: 874 Days
• Field of Search: 395/24, 395/22, 395/25, 395/27
• US Class Current: 706/26
• CPC Class Codes: G06N 3/063 using electronic means

Abstract

A neural network is disclosed in which communication between processing elements occurs by radio waves in a waveguide. Radio wave communication using common carrier signals by transceivers in a waveguide allows processing elements to communicate wirelessly and simultaneously. Each processing element includes a radio frequency transceiver and an accompanying antenna which performs the neuron summing operation because input signals simultaneously received from plural processing elements by the antenna add. The weights on each input are provided by different spatial relationships between the transmitting processing elements and the receiving processing element which causes signal strength loses through the waveguide to be different. Each receiving processing element performs a neural threshold or sigmoid operation on the summed signal received from the transceiver and then a strength (amplitude scaling) can be applied to the output before the processing element transmits that output to the other processing elements in the system. Processing elements are grouped, allowing one group to transmit while the other group is receiving. Wafer scale electronics including transceivers and analog processing elements are combined with a comparably sized waveguide to produce a compact device.

Claims

1. A neural network, comprising:

processing elements performing neural transfer function operations; and

a waveguide for simultaneously connecting said processing elements to each other.

2. A neural network, comprising:

processing elements performing neural transfer function operations on input signals; and

a waveguide for summing the input signals for the processing elements.

3. A neural network, comprising:

processing elements performing neural transfer function operations on input signals; and

waveguide means for weighting the input signals.

4. A neural processing element communicating with other processing elements, comprising:

a waveguide and an antenna positioned in said waveguide and wirelessly, simultaneously receiving and summing signals from the other processing elements; and

transfer function means for performing a neural transfer function on the summed signals.

5. A neural processing element as recited in claim 4, wherein the neural transfer function comprises a non-linear transfer function.

6. A neural processing element as recited in claim 4, wherein said waveguide comprises means for reflecting and/or absorbing radio waves.

7. A neural network, comprising:

wafer scale analog electronics including neurons connected to radio transceivers;

antennas connected to the transceivers; and

a waveguide surrounding said antennas.

8. A parallel processing neural network comprising:

first through fourth layers of neural processing elements each having a processing cycle with phases, the phases overlapping and the layers processing in parallel; and

radio wave means for communicating between said layers.

9. A neural processing element communicating with other processing elements through an adjacent waveguide, comprising:

an antenna projecting into the waveguide and receiving radio frequency signals from the other processing elements;

a transceiver connected to said antenna; and

a neural processing element transfer function unit connected to said transceiver.

10. A neural network, comprising:

a waveguide; and

neural network processing elements each for producing an output signal in said waveguide and each comprising;

summing means for wirelessly summing the output signals; and

transfer function means for performing a neural transfer function responsive to the summed output signals.

11. A network as recited in claim 10, wherein said processing elements are spatially separated, thereby applying spatially determined weights to the output signals.

12. A network recited in claim 11, wherein the spatial separation is irregular.

13. A network as recited in claim 10, further comprising means for applying a strength to an output of said transfer function means.

14. A neural network, comprising:

a waveguide;

antennas irregularly, spatially positioned in said waveguide;

transceivers corresponding to and connected to said antennas; and

neural processing element transfer function units corresponding to and connected to said transceivers.

15. A network as recited in claim 14, wherein said antennas, transceivers and units are divided into first and second groups and said first group transmits while said second group receives.

16. A method of performing neural network operations between first and second groups of neural processing elements, comprising the steps of:

(a) performing, by the first group, a first neural processing element transfer function to produce first neuron signals;

(b) radio frequency transmitting the first neuron signals;

(c) guiding the transmitted first neuron signals from the first group to the second group; and

(d) performing, by the second group, a second neural processing element transfer function on the first neuron signals.

17. A method as recited in claim 16, wherein step (c) is performed in a waveguide.

18. A method as recited in claim 16 further comprising the steps of:

(d) producing, by the second group, second neuron signals;

(e) radio frequency transmitting the second neuron signals from the second group to the first group; and

(f) performing, by the first group, a third neural processing element transfer function on the second neuron signals.