Patent US5943659A successfully refreshed from Google Patents.
US5943659A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Deterministic encoding of fuzzy finite state automata in continuous recurrent neural networks
Assignee: Rakuten Group Inc
•
Filing: 1995-10-03
•
Granted: 1999-08-24
•
Expires: 2015-10-03
Abstract
Abstract Based on the encoding of deterministic finite-state automata (DFA) in discrete-time, second-order recurrent neural networks, an algorithm constructs an augmented recurrent neural network that encodes a FFA and recognizes a given fuzzy regular language with arbitrary accuracy.
Inventors
- 👤 C. Lee Giles
- 👤 Christian Walter Omlin
- 👤 Karvel Kuhn Thornber
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-08 06:42
CPC Classifications
G; G05; G05B; G05B13/00; G05B13/02; G05B13/0265; G05B13/0285; G06; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/043
Claims Analysis
Total Claims: 6
Claim 1
Independent Claim
144 words
A fuzzy finite state automata in continuous recurrent neural networks, the fuzzy finite state automata providing an output control signal for controlling a controlled system without requiring learning, the fuzzy finite state automata comprising: input neurons in a recurrent neural network receiving input signals and providing strings; sigmoidal discriminators receiving said strings and computing and providing as outputs the current state of the network; recurrent state neurons providing as outputs a weighted sum of all the outputs of the sigmoidal discriminators; feedback delay providing the outputs of said recurrent state neurons as time delayed inputs to said sigmoidal discriminators; fuzzy membership weights receiving the output of said recurrent state neurons and assigning fuzzy weights to each said output so that learning is not required; and network output neurons summing the fuzzy weighted outputs for providing the output control signal for controlling the controlled system.
Claim 2
Dependent Claim
144 words
A fuzzy finite state automata in continuous recurrent neural networks, the fuzzy finite state automata providing an output control signal for controlling a controlled system without requiring learning, the fuzzy finite state automata comprising: input neurons in a recurrent neural network receiving input signals and providing strings; sigmoidal discriminators receiving said strings and computing and providing as outputs the current state of the network; recurrent state neurons providing as outputs a weighted sum of all the outputs of the sigmoidal discriminators; feedback delay providing the outputs of said recurrent state neurons as time delayed inputs to said sigmoidal discriminators; fuzzy membership weights receiving the output of said recurrent state neurons and assigning fuzzy weights to each said output so that learning is not required; and network output neurons summing the fuzzy weighted outputs for providing the output control signal for controlling the controlled system.
Claim 3
Dependent Claim
104 words
A fuzzy finite state automata in continuous recurrent neural networks as set forth in claim 1, where said sigmoidal discriminator computes the function ##EQU8## and ##EQU9## where b i is the bias associated with hidden recurrent state neurons S i , I k is the input neuron for symbol a k , the product S j .sup.(t) I k .sup.(t) corresponds to the state transition δ(q i , q k )=q i , and W ijk is +H if δ(q i ,q k )=q i and -H if δ(q i ,q k )≠q i where H is a weight strength of an output neuron.
Claim 4
Independent Claim
136 words
A method of generating an output control signal from a fuzzy finite state automata in a continuous recurrent neural network for controlling a controlled system without requiring learning, the method comprising the steps of: receiving input signals commensurate with a measured value and providing output strings responsive thereto; subjecting the outputs strings to a sigmoidal discriminant function; providing first signals commensurate with a weighted sum of all output strings subjected to the sigmoidal discriminant function and time delayed signals which were subjected to the sigmoidal discriminant function; feeding back said time delayed signals as time delayed signals to be combined with the received input signals; subjecting said first signals to fuzzy membership weighting so that learning is not required; and summing said fuzzy weighted signals to generate the output control signal for controlling the controlled system.
Claim 5
Independent Claim
136 words
A method of generating an output control signal from a fuzzy finite state automata in a continuous recurrent neural network for controlling a controlled system without requiring learning, the method comprising the steps of: receiving input signals commensurate with a measured value and providing output strings responsive thereto; subjecting the outputs strings to a sigmoidal discriminant function; providing first signals commensurate with a weighted sum of all output strings subjected to the sigmoidal discriminant function and time delayed signals which were subjected to the sigmoidal discriminant function; feeding back said time delayed signals as time delayed signals to be combined with the received input signals; subjecting said first signals to fuzzy membership weighting so that learning is not required; and summing said fuzzy weighted signals to generate the output control signal for controlling the controlled system.
Claim 6
Dependent Claim
120 words
A method of generating an output signal in a fuzzy finite state automata in a continuous recurrent neural network as set forth in claim 3, where said sigmoidal discriminant function is ##EQU10## and ##EQU11## where b i is the bias associated with hidden recurrent state neurons S i , I k is the input neuron for symbol a k , the product S j .sup.(t) I k .sup.(t) corresponds to the state transition δ(q i ,q k )=q i and W ijk is +H if δ(q i , q k )=q i and -H if δ(q i , q h )≠q i ! δ(q i , q k )≠q i where H is a weight strength of an output neuron.
References Cited (Backward Citations)
US5168549A
Cited
JPH02292602A
Cited
US5471381A
Cited
US5606646A
Cited
US5524176A
Cited
US5761384A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2002-08-23 | 2003-02-22 | 2003-08-24 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2006-08-23 | 2007-02-22 | 2007-08-24 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2010-08-23 | 2011-02-22 | 2011-08-24 | $7700.00 | Window Closed / Lapsed |