Patent US4904881A successfully refreshed from Google Patents.
US4904881A
Expired / Lapsed
US
⏱️ 0.0 years remaining
EXCLUSIVE-OR cell for neural network and the like
Assignee: Intel Corp
•
Filing: 1989-02-10
•
Granted: 1990-02-27
•
Expires: 2009-02-10
Abstract
Abstract A semiconductor cell for producing an output current that is related to the match between an input vector pattern and a weighting pattern is described. The cell is particularly useful as a synapse cell within a neural network to perform pattern recognition tasks. The cell includes a pair of input lines for receiving a differential input vector element value and a pair of output lines for providing a difference current to a current summing neural amplifier. A plurality of floating gate devices each having a floating gate member are employed in the synapse cell to store charge in accordance with a predetermined weight pattern. Each of the floating gate devices is uniquely coupled to a combination of an output current line and an input voltage line such that the difference current provided to the neural amplifier is related to the match between the input vector and the stored weight.
Inventors
- 👤 Hernan A. Castro
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-27 16:55
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/06; G06N3/063; G06N3/065; G06F; G06F18/00; G06F18/20; G06F18/22
Claims Analysis
Total Claims: 18
Claim 1
Independent Claim
81 words
A semiconductor cell for producing an output current which is related to the match between an input pattern and a weight pattern, said cell comprising: circuit means for receiving said input pattern and producing said output current, said circuit means comprising a plurality of memory devices for storing said weight pattern such that said output current is related to the match between said input pattern and said weight pattern; and programming means for programming said weight pattern into said circuit means.
Claim 2
Dependent Claim
81 words
A semiconductor cell for producing an output current which is related to the match between an input pattern and a weight pattern, said cell comprising: circuit means for receiving said input pattern and producing said output current, said circuit means comprising a plurality of memory devices for storing said weight pattern such that said output current is related to the match between said input pattern and said weight pattern; and programming means for programming said weight pattern into said circuit means.
Claim 3
Dependent Claim
14 words
The semiconductor cell of claim 1 wherein said memory devices are floating gate devices.
Claim 4
Dependent Claim
34 words
The semiconductor cell according to claim 2 wherein said input pattern is coupled to the gates of said floating gate devices and said output current is supplied to the input of a neural amplifier.
Claim 5
Dependent Claim
91 words
A semiconductor cell for use in computing the Hamming distance between a weight pattern and an input pattern, said cell comprising: a plurality of floating gate devices for storing said weight pattern, each of said floating gate devices having a floating gate, a drain coupled to an output and a gate coupled to said input pattern, said cell providing a current at said output which is related to the Hamming distance between said input pattern and said weight pattern; and means for programming said weight pattern into said floating gate devices.
Claim 6
Dependent Claim
91 words
A semiconductor cell for use in computing the Hamming distance between a weight pattern and an input pattern, said cell comprising: a plurality of floating gate devices for storing said weight pattern, each of said floating gate devices having a floating gate, a drain coupled to an output and a gate coupled to said input pattern, said cell providing a current at said output which is related to the Hamming distance between said input pattern and said weight pattern; and means for programming said weight pattern into said floating gate devices.
Claim 7
Dependent Claim
37 words
The cell of claim 4 wherein said input pattern and said weight pattern represent binary values such that said current provided at said output corresponds to the exclusive-OR function of said input pattern and said weight pattern.
Claim 8
Dependent Claim
22 words
The cell of claim 4 wherein said current is proportional to an analog multiplication of said input pattern and said weight pattern.
Claim 9
Dependent Claim
134 words
In a neural network, a synapse cell useful in performing pattern recognition tasks comprising: a pair of input lines for receiving a differential input vector value; a pair of output lines for providing a difference current; and first, second, third and fourth floating gate devices, said first and said second floating gate devices having their gates coupled to one of said input lines and their drains coupled to alternate ones of said output lines, said third and fourth floating gate devices having their gates coupled to another of said input lines and their drains also coupled to alternate ones of said output lines, said floating gate devices storing a weight such that said difference current provided across said pair of output lines is proportional to the match between said input vector and said weight.
Claim 10
Dependent Claim
134 words
In a neural network, a synapse cell useful in performing pattern recognition tasks comprising: a pair of input lines for receiving a differential input vector value; a pair of output lines for providing a difference current; and first, second, third and fourth floating gate devices, said first and said second floating gate devices having their gates coupled to one of said input lines and their drains coupled to alternate ones of said output lines, said third and fourth floating gate devices having their gates coupled to another of said input lines and their drains also coupled to alternate ones of said output lines, said floating gate devices storing a weight such that said difference current provided across said pair of output lines is proportional to the match between said input vector and said weight.
Claim 11
Dependent Claim
19 words
The synapse cell of claim 7 further comprising programming means for programming said weight into said floating gate devices.
Claim 12
Dependent Claim
32 words
The synapse cell of claim 8 wherein said input vector and said weight represent binary values such that said difference current corresponds to the exclusive-OR function of said vector and said weight.
Claim 13
Dependent Claim
22 words
The synapse cell according to claim 9 wherein said pair of output lines are coupled to the inputs of a neural amplifier.
Claim 14
Dependent Claim
22 words
The synapse cell of claim 7 wherein said current difference is proportional to analog multiplication of said input vector and said weight.
Claim 15
Dependent Claim
142 words
In a neural network, a synapse cell for generating an analog current related to the binary dot product of an input vector and a weight, said cell comprising: first and second input lines providing said input vector and its complement, respectively; a pair of output lines providing said analog current: a pair of floating gate devices each having their gates coupled to said first line, their drains coupled to alternate ones of said output lines, and their sources grounded; another pair of floating gate devices each having their gates coupled to said second input line, their drains coupled to alternate ones of said output lines and their sources grounded, each of said floating gate devices also having a floating gate for storing said weight such that said analog current difference is related to the match between said weight and said input vector.
Claim 16
Dependent Claim
142 words
In a neural network, a synapse cell for generating an analog current related to the binary dot product of an input vector and a weight, said cell comprising: first and second input lines providing said input vector and its complement, respectively; a pair of output lines providing said analog current: a pair of floating gate devices each having their gates coupled to said first line, their drains coupled to alternate ones of said output lines, and their sources grounded; another pair of floating gate devices each having their gates coupled to said second input line, their drains coupled to alternate ones of said output lines and their sources grounded, each of said floating gate devices also having a floating gate for storing said weight such that said analog current difference is related to the match between said weight and said input vector.
Claim 17
Dependent Claim
18 words
The cell according to claim 12 further comprising programming means for programming said weight onto said floating gates.
Claim 18
Dependent Claim
23 words
The cell according to claim 13 wherein said pair of output lines is coupled to the inputs of a current summing neural amplifier.
References Cited (Backward Citations)
US3546601A
Cited
US4027175A
Cited
US4055773A
Cited
US4453096A
Cited
US4202044A
Cited
US4617479B1
Cited
US4617479A
Cited
US4660166A
Cited
US4695979A
Cited
US4760437A
Cited
US4773024A
Cited
US4802103A
Cited
US4797625A
Cited
US4782460A
Cited
US4807168A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1993-02-26 | 1993-08-28 | 1994-02-27 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 1997-02-26 | 1997-08-28 | 1998-02-27 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2001-02-26 | 2001-08-28 | 2002-02-27 | $7700.00 | Window Closed / Lapsed |