US5099434A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Continuous-time optical neural network
Assignee: Northrop Grumman Corp
•
Filing: 1990-08-03
•
Granted: 1992-03-24
•
Expires: 2010-08-03
Abstract
Abstract An all-optical, continuous-time, recurrent neural network is disclosed which is capable of executing a broad class of energy-minimizing neural net algorithms. The network is a resonator which contains a saturable, two-beam amplifier; two volume holograms; and a linear, two-beam amplifier. The saturable amplifier permits, through the use of a spatially patterned signal beam, the realization of a two-dimensional optical neuron array; the two volume holograms provide adaptive, global network interconnectivity; and the linear amplifier supplies sufficient resonator gain to permit convergent operation of the network.
Inventors
- 👤 Harold M. Stoll
- 👤 Li-Shing Lee
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-02 14:15
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/06; G06N3/067; G06E; G06E3/00; G06E3/001; G06V; G06V10/00; G06V10/88; G03; G03H; G03H1/00; G03H1/26; G03H1/2645; G03H1/265; G03H1/0005; G03H2001/0066; G03H2260/00; G03H2260/50; G03H2260/54
Claims Analysis
Total Claims: 23
Claim 1
Independent Claim
120 words
An optical neural network for finding at least one minimum of a system energy function comprising means for forming an array of saturable optical amplifiers into an optical neuron array having inputs and outputs, means for forming an adaptive and bipolar global optical interconnect for multiplying the outputs of said optical neuron array by an interconnect matrix and for optically interconnecting the result to the inputs of said array to form a recurrent, resonant network, optical input means for receiving an input signal and for structuring the interconnect matrix from an inputted training set, and means for supplying gain to achieve convergent, resonant operation of said network so that operation of said optical neural network finds system energy function minima.
Claim 2
Dependent Claim
120 words
An optical neural network for finding at least one minimum of a system energy function comprising means for forming an array of saturable optical amplifiers into an optical neuron array having inputs and outputs, means for forming an adaptive and bipolar global optical interconnect for multiplying the outputs of said optical neuron array by an interconnect matrix and for optically interconnecting the result to the inputs of said array to form a recurrent, resonant network, optical input means for receiving an input signal and for structuring the interconnect matrix from an inputted training set, and means for supplying gain to achieve convergent, resonant operation of said network so that operation of said optical neural network finds system energy function minima.
Claim 3
Dependent Claim
21 words
An optical neural network of claim 1 wherein: said optical interconnect means includes holographic interconnect means for interconnecting said optical neurons.
Claim 4
Dependent Claim
16 words
An optical neural network of claim 1 wherein: said optical input means comprises an optical beamsplitter.
Claim 5
Dependent Claim
21 words
An optical neural network of claim 1 wherein: said interconnect means includes adaptive interconnect means for adaptively interconnecting said optical neurons.
Claim 6
Dependent Claim
24 words
An optical neural network of claim 1 wherein: said interconnect means includes photorefractive interconnect means for interconnecting said optical neurons in a photorefractive material.
Claim 7
Dependent Claim
22 words
An optical neural network of claim 5 wherein: said interconnect means includes first and second photorefractive crystals independently located in said resonator.
Claim 8
Dependent Claim
25 words
An optical neural network of claim 1 wherein: said interconnect means includes volume hologram means for forming at least one volume hologram in said resonator.
Claim 9
Dependent Claim
23 words
An optical neural network of claim 1 wherein: said gain means includes means for supplying both linear and nonlinear gain in said network.
Claim 10
Dependent Claim
29 words
An optical neural network of claim 8 wherein: said nonlinear gain means includes at least one photorefractive crystal and means for supplying a coherent pump beam to said crystal.
Claim 11
Dependent Claim
32 words
An optical neural network of claim 9 wherein: said pump beam intersects one of said photorefractive crystals at an angle so as to effectively only intersect one neural volume within said crystal.
Claim 12
Dependent Claim
43 words
An optical neural network of claim 9 wherein: said gain means includes two photorefractive crystals each having a means for supplying a coherent pump beam to the respective crystals and where the coherent pump beams intersect the respective crystals at different effective angles.
Claim 13
Dependent Claim
29 words
An optical neural network of claim 8 wherein: said nonlinear gain means includes means for supplying a coherent pump beam which only intersects one neural volume in said network.
Claim 14
Dependent Claim
25 words
An optical neural network of claim 1 wherein: said interconnect means includes Fourier-space and object-space hologram means for holographically interconnecting optical neurons in said network.
Claim 15
Dependent Claim
24 words
An optical neural network of claim 12 further including: means for generating a common reference beam for said Fourier-space and said object-space hologram means.
Claim 16
Dependent Claim
33 words
An optical neural network of claim 1 further including: phase encoding means operatively associated in said network with said neuron interconnect means for supplying a phase-encoded optical signal to said neuron interconnect means.
Claim 17
Dependent Claim
68 words
An optical neural network which comprises: a ring resonator including a saturable amplifier, first and second volume holograms, and a linear amplifier, said saturable amplifier in association with a spatially patterned input signal beam forming a two-dimensional optical neuron array, said first and second volume holograms forming an adaptive and bipolar global network interconnect, and said linear amplifier for supplying gain for resonant convergent operation of said network.
Claim 18
Dependent Claim
68 words
An optical neural network which comprises: a ring resonator including a saturable amplifier, first and second volume holograms, and a linear amplifier, said saturable amplifier in association with a spatially patterned input signal beam forming a two-dimensional optical neuron array, said first and second volume holograms forming an adaptive and bipolar global network interconnect, and said linear amplifier for supplying gain for resonant convergent operation of said network.
Claim 19
Dependent Claim
23 words
An optical neural network of claim 16 wherein: said first hologram comprises a Fourier-space hologram, and said second hologram comprises an object-space hologram.
Claim 20
Dependent Claim
33 words
An optical neural network of claim 16 further including: a phase encoder operatively associated in said network with said first and second holograms for supplying a phase-encoded optical signal to said first hologram.
Claim 21
Dependent Claim
21 words
An optical neural network of claim 16 wherein: said linear amplifier and said saturable amplifier each comprise two-beam photorefractive optical amplifiers.
Claim 22
Dependent Claim
35 words
An optical neural network of claim 16 wherein: said linear amplifier and said saturable amplifier each comprise a photorefractive barium titanate crystal, and said first and second holograms each comprise a photorefractive lithium niobate crystal.
Claim 23
Dependent Claim
31 words
The optical neural network as in claim 16 further including means for inputting a stimulus pattern into said resonator with a tilt incorporated into the phase front of said stimulus pattern.
References Cited (Backward Citations)
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1995-03-24 | 1995-09-23 | 1996-03-24 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1999-03-24 | 1999-09-23 | 2000-03-24 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2003-03-24 | 2003-09-23 | 2004-03-24 | $7700.00 | Window Closed / Lapsed |