Patent US5065339A successfully refreshed from Google Patents.
US5065339A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Orthogonal row-column neural processor
Assignee: International Business Machines Corp
•
Filing: 1990-05-22
•
Granted: 1991-11-12
•
Expires: 2010-05-22
Abstract
Abstract The neural computing paradigm is characterized as a dynamic and highly parallel computationally intensive system typically consisting of input weight multiplications, product summation, neural state calculations, and complete connectivity among the neurons. Herein is described neural network architecture called SNAP which uses a unique intercommunication scheme within an array structure that provides high performance for completely connected network models such as the Hopfield model. SNAP's packaging and expansion capabilities are addressed, demonstrating SNAP's scalability to larger networks. Each neuron generating a neuron value from a selected set of input function elements and communicating said neuron value back to said set of input function elements. The total connectivity of each neuron to all neurons is accomplished by an orthogonal row-column relationship of neurons where a given multiplier element operates during a first cycle as a row element within an input function to a column neuron, and during a second cycle as a column element within an input function to a row neuron.
Inventors
- 👤 Stamatis Vassiliadis
- 👤 Gerald G. Pechanek
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-05 19:20
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/06; G06N3/063; G06N3/10
Claims Analysis
Total Claims: 22
Claim 1
Independent Claim
52 words
A computing apparatus, comprising a plurality of input function elements, each input function element selectively allocated to orthogonal neurons, each neuron including means for generating a neuron value from a selected set of input function elements and means for communicating said neuron value back to said selected set of input function elements.
Claim 2
Dependent Claim
52 words
A computing apparatus, comprising a plurality of input function elements, each input function element selectively allocated to orthogonal neurons, each neuron including means for generating a neuron value from a selected set of input function elements and means for communicating said neuron value back to said selected set of input function elements.
Claim 3
Dependent Claim
28 words
The computing apparatus of claim 1 wherein said input function elements include means for applying a weight value to an input value to generate a weighted input value.
Claim 4
Dependent Claim
15 words
The computing apparatus of claim 2 wherein said means for applying comprises a digital multiplier.
Claim 5
Dependent Claim
51 words
The computing apparatus of claim 2 wherein said means for generating a neuron value includes means for combining the weighted input values from each of its input function elements and activation function means for applying an activation function to the output of said means for combining to generate said neuron value.
Claim 6
Dependent Claim
15 words
The computing apparatus of claim 4 wherein said means for combining comprises an adder tree.
Claim 7
Dependent Claim
14 words
The computing apparatus of claim 4 wherein said activation function comprises a sigmoid generator.
Claim 8
Dependent Claim
34 words
The computing apparatus of claim 4 wherein said means for communicating includes communication path means selectively operable for communicating said neuron value and for communicating said weighted input value to said activation function means.
Claim 9
Dependent Claim
24 words
The computing apparatus of claim 7 wherein said communication path means comprises separate transmission paths for said weighted input value and said neuron value.
Claim 10
Dependent Claim
23 words
The computing apparatus of claim 4 wherein said means for combining includes iterative adder means for accumulating partial sums of weighted input values.
Claim 11
Dependent Claim
23 words
The computing apparatus of claim 4 including means for operating said means for generating and said means for communicating in bit serial mode.
Claim 12
Dependent Claim
33 words
The computing apparatus of claim 1 including means for allocating each said input function element selectively as a row element in a column neuron and as a column element in a row neuron.
Claim 13
Independent Claim
72 words
A method for operating an array processor comprising orthogonal sets of neurons and a plurality of input function elements, the method comprising the steps of (1) operating a first neuron upon a first subset of said input functions to generate and load back into said first subset a neuron value, and (2) allocating each of said first subset of input function elements to a different one of a set of orthogonal neurons.
Claim 14
Independent Claim
72 words
A method for operating an array processor comprising orthogonal sets of neurons and a plurality of input function elements, the method comprising the steps of (1) operating a first neuron upon a first subset of said input functions to generate and load back into said first subset a neuron value, and (2) allocating each of said first subset of input function elements to a different one of a set of orthogonal neurons.
Claim 15
Dependent Claim
60 words
The computing apparatus of claim 1 wherein said orthogonal neurons comprise a plurality N (where "N" is an integer) of row neurons and a plurality N (where "N" is an integer) of column neurons, with each input function element selectively allocated as a row input function in a column neuron and as a column input function in a row neuron.
Claim 16
Dependent Claim
49 words
The computing apparatus of claim 13 further including means for operating at least two row neurons on N (where "N" is an integer) column input function elements in parallel to produce partial column neuron values, and means for accumulating said partial column neuron values to generate column neuron values.
Claim 17
Dependent Claim
34 words
The computing apparatus of claim 13 further including means for operating at least two row neurons on N (where "N" is an integer) column input function elements in parallel to produce row neuron values.
Claim 18
Dependent Claim
49 words
The computing apparatus of claim 13 further including means for operating at least two column neurons on N (where "N" is an integer) row input function elements in parallel to produce partial row neuron values, and means for accumulating said partial row neuron values to generate row neuron values.
Claim 19
Dependent Claim
33 words
The computing apparatus of claim 13 further including means for operating at least two column neurons on N (where "N" an integer) row input function elements in parallel to produce column neuron values.
Claim 20
Dependent Claim
24 words
The computing apparatus of claim 1 wherein said orthogonal neurons comprise a plurality of planes of operating as equivalent to row or column neurons.
Claim 21
Dependent Claim
16 words
The computing apparatus of claim 13 wherein said row neurons are organized in a planar structure.
Claim 22
Dependent Claim
16 words
The computing apparatus of claim 13 wherein said column neurons are organized in a planar structure.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1994-11-11 | 1995-05-13 | 1995-11-12 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1998-11-11 | 1999-05-13 | 1999-11-12 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2002-11-11 | 2003-05-13 | 2003-11-12 | $7700.00 | Window Closed / Lapsed |