Patent US5434951A successfully refreshed from Google Patents.
US5434951A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Neural network system having minimum energy function value
Assignee: Toshiba Corp
•
Filing: 1993-07-22
•
Granted: 1995-07-18
•
Expires: 2013-07-22
Abstract
Abstract N neural networks having different set-values are provided, where N is an integer greater than 2. Each neural network has plurality of artificial neurons and processes information. An optimal output detecting circuit receives outputs of the N neural networks and determines the optimal one of the neural networks based on the outputs of the N neural networks. An output circuit receives and outputs the output of the neural network detected by the optimal output detecting circuit.
Inventors
- 👤 Ryuichi Kuwata
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 15:24
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/045; G06N3/06; G06N3/063
Claims Analysis
Total Claims: 18
Claim 1
Independent Claim
174 words
A neural network system comprising: N neural networks having different set-values which define operations of the N neural networks, said set-values being arbitrarily settable, where N is an integer greater than 1, each having a plurality of artificial neurons arranged in rows and columns, an output of each artificial neuron being fed-back to other artificial neurons of the same row and column, said N neural networks operating in parallel; first means, connected to said N neural networks, for receiving outputs of said N neural networks, obtaining energy function values which indicate energy states of the N neural networks, comparing the energy function values, and determining a minimum energy function value from among the energy function values based on the outputs of the N neural networks; and output means, connected to said N neural networks and said first means, for receiving a signal from said first means, said signal indicating a first neural network having said minimum energy function value, and outputting an output of said first neural network having the minimum energy function value.
Claim 2
Dependent Claim
174 words
A neural network system comprising: N neural networks having different set-values which define operations of the N neural networks, said set-values being arbitrarily settable, where N is an integer greater than 1, each having a plurality of artificial neurons arranged in rows and columns, an output of each artificial neuron being fed-back to other artificial neurons of the same row and column, said N neural networks operating in parallel; first means, connected to said N neural networks, for receiving outputs of said N neural networks, obtaining energy function values which indicate energy states of the N neural networks, comparing the energy function values, and determining a minimum energy function value from among the energy function values based on the outputs of the N neural networks; and output means, connected to said N neural networks and said first means, for receiving a signal from said first means, said signal indicating a first neural network having said minimum energy function value, and outputting an output of said first neural network having the minimum energy function value.
Claim 3
Dependent Claim
25 words
The neural network system according to claim 1, wherein said first means includes means for receiving each of the outputs of said N neural networks.
Claim 4
Dependent Claim
52 words
The neural network system according to claim 1, wherein said first means includes re-setting means, connected to said N neural networks, for re-setting the set values of said N neural networks and operating said N neural networks again after said first means determines said first neural network having said minimum energy value.
Claim 5
Dependent Claim
153 words
The neural network system according to claim 1, wherein said first means includes: first setting means, connected to said N neural networks, for setting said N neural networks at different set-values after said first means determines said first neural network having said minimum energy value, and means for repetitively operating said N neural networks M times, where M is a positive integer greater than 1; second means for obtaining and storing an energy function value of each of said N neural networks; means for selecting P (P being a positive integer) of N×M energy function values obtained by and stored in said second means, which are less than any other energy function values; and second setting means for setting set-values of said N neural networks and for operating said N neural networks, said set-values having values similar to those set to said N neural networks when the P energy function values are obtained.
Claim 6
Dependent Claim
45 words
A neural network system according to claim 1, wherein each of said neural networks comprises a plurality of artificial neurons connected to one another; said set-values are at least one of a bias value, an output coefficient, and an initial value of a calculation amplifier.
Claim 7
Dependent Claim
20 words
A neural network system according to claim 1, wherein said N neural networks have a same structure and different set-values.
Claim 8
Dependent Claim
145 words
A neural network comprising: a plurality of artificial neurons each having a respective readiness for activating arranged in rows and columns; connecting means for connecting said plurality of artificial neurons to one another in an inhibitory synaptic manner so that only one artificial neuron is activated in one row and one column, the connecting means connecting a neuron to every other neuron in a same row and column; and means for dividing said plurality of artificial neurons into hierarchical groups; wherein the artificial neurons of each hierarchical group have a readiness for activating which is different from that of any other hierarchical group; and wherein said hierarchical groups are one of rows of artificial neurons, columns of artificial neurons, and groups each consisting of artificial neurons having the same (i+j) value, where i is a row serial number, and j is a column serial number.
Claim 9
Dependent Claim
145 words
A neural network comprising: a plurality of artificial neurons each having a respective readiness for activating arranged in rows and columns; connecting means for connecting said plurality of artificial neurons to one another in an inhibitory synaptic manner so that only one artificial neuron is activated in one row and one column, the connecting means connecting a neuron to every other neuron in a same row and column; and means for dividing said plurality of artificial neurons into hierarchical groups; wherein the artificial neurons of each hierarchical group have a readiness for activating which is different from that of any other hierarchical group; and wherein said hierarchical groups are one of rows of artificial neurons, columns of artificial neurons, and groups each consisting of artificial neurons having the same (i+j) value, where i is a row serial number, and j is a column serial number.
Claim 10
Dependent Claim
48 words
The neural network according to claim 7, wherein said setting means sets the readiness for activating the artificial neurons of each hierarchical group differently by setting a bias signal supplied to the artificial neurons, a threshold value thereof, the output function thereof, or inhibitory connection coefficient thereof differently.
Claim 11
Dependent Claim
53 words
The neural network according to claim 7, further comprising said hierarchical groups being rows or columns of artificial neurons, and input means for inputting an input signal to each of the artificial neurons of the same column or row; said input means inputting input signals to artificial neurons in every row and column.
Claim 12
Dependent Claim
60 words
The neural network according to claim 7, wherein means for assigning the rows or columns of artificial neurons to events, respectively, and assigning the columns or rows of artificial neurons are assigned to the priorities of the events, so that an order in which the events should take place is determined by processing signals output from all the artificial neurons.
Claim 13
Dependent Claim
95 words
The neural network according to claim 10, further comprising: first means for determining whether or not there is an event which will not satisfy a predetermined condition according to results of processing of signals of the artificial neurons; parameter-setting means for changing the activating readiness of the artificial neurons related to an event, if any, which will not satisfy the predetermined condition; and second means for detecting that all events satisfy the predetermined condition and for repeatedly controlling said artificial neurons, said first means, and said parameter-setting means until all events satisfy the predetermined condition.
Claim 14
Dependent Claim
104 words
The neural network according to claim 10, further comprising: first means for determining whether or not there is an event which does not satisfy a predetermined condition according to results of processing of signals of the artificial neurons; parameter-setting means for replacing the row of artificial neurons related to an event, if any, which does not satisfy the predetermined condition with another row whose serial number is less than said rows; and second means for detecting that all events satisfy the predetermined condition and for repeatedly controlling said artificial neurons, said first means, and said parameter-setting means until all events satisfy the predetermined condition.
Claim 15
Dependent Claim
126 words
A neural network system comprising: N neural networks having different operation characteristics, where N is an integer greater than 1, each having a plurality of artificial neurons arranged in rows and columns, said N neural networks operating in parallel; first means, connected to said neural networks, for receiving outputs of said N neural networks, and calculating energy function values of the N neural networks; second means, connected to the first means, for comparing the energy function values, and selecting a neural network having a minimum energy function value; and third means, connected to said N neural networks and said second means, for receiving said outputs of said N neural networks and for outputting an output value of the neural network having the minimum energy function value.
Claim 16
Dependent Claim
126 words
A neural network system comprising: N neural networks having different operation characteristics, where N is an integer greater than 1, each having a plurality of artificial neurons arranged in rows and columns, said N neural networks operating in parallel; first means, connected to said neural networks, for receiving outputs of said N neural networks, and calculating energy function values of the N neural networks; second means, connected to the first means, for comparing the energy function values, and selecting a neural network having a minimum energy function value; and third means, connected to said N neural networks and said second means, for receiving said outputs of said N neural networks and for outputting an output value of the neural network having the minimum energy function value.
Claim 17
Dependent Claim
123 words
A neural network comprising: a plurality of artificial neurons arranged in rows and columns, each artificial neuron having a respective readiness for activating to form hierarchical groups; and connecting means for connecting said plurality of artificial neurons to one another in an inhibitory synaptic manner so that only one artificial neuron is activated in one row and one column; wherein the artificial neurons of each hierarchical group have a readiness for activating which is different from that of any other hierarchical group; and wherein said hierarchical groups are one of rows of artificial neurons, columns of artificial neurons, and groups each including artificial neurons having the same (i+j) value, where i is a row serial number, and j is a column serial number.
Claim 18
Dependent Claim
123 words
A neural network comprising: a plurality of artificial neurons arranged in rows and columns, each artificial neuron having a respective readiness for activating to form hierarchical groups; and connecting means for connecting said plurality of artificial neurons to one another in an inhibitory synaptic manner so that only one artificial neuron is activated in one row and one column; wherein the artificial neurons of each hierarchical group have a readiness for activating which is different from that of any other hierarchical group; and wherein said hierarchical groups are one of rows of artificial neurons, columns of artificial neurons, and groups each including artificial neurons having the same (i+j) value, where i is a row serial number, and j is a column serial number.
References Cited (Backward Citations)
US3311895A
Cited
US3284780A
Cited
US4660166A
Cited
US4760604A
Cited
US4796199A
Cited
US4782460A
Cited
US4858147A
Cited
US4942517A
Cited
JPH01147657A
Cited
US5285522A
Cited
US4849925A
Cited
JPH01189753A
Cited
US4931674A
Cited
US5170463A
Cited
US5075871A
Cited
US4972363A
Cited
US4988891A
Cited
US5075868A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 1998-07-17 | 1999-01-16 | 1999-07-18 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2002-07-17 | 2003-01-16 | 2003-07-18 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2006-07-17 | 2007-01-16 | 2007-07-18 | $7700.00 | Window Closed / Lapsed |