Patent US5255347A successfully refreshed from Google Patents.
US5255347A
Expired / Lapsed
US
⏱️ 0.0 years remaining
Neural network with learning function
Assignee: Hitachi Ltd
•
Filing: 1991-04-23
•
Granted: 1993-10-19
•
Expires: 2011-04-23
Abstract
Abstract A neural network system capable of performing integrated processing of a plurality of information includes a feature extractor group for extracting a plurality of learning feature data from learning data in a learning mode and a plurality of object feature data from object data to be processed in an execution mode, and an information processing unit for learning features of the learning data, based on the plurality of learning feature data from the feature extractor group and corresponding teacher data in the learning mode, and determining final learning result data from the plurality of object feature data from the feature extractor group in accordance with the learning result, including a logic representing relation between the plurality of object feature data in the execution mode.
Inventors
- 👤 Ikuo Matsuba
- 👤 Ichirou Sugita
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-10-28 19:41
CPC Classifications
G; G06; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/045; G06N3/044; G06N3/0442; G06N3/08; G06N3/09
Claims Analysis
Total Claims: 9
Claim 1
Independent Claim
247 words
A neural network system comprising: feature extracting means for i) extracting N heterogeneous learning feature data, where N is an integer greater than one, from each of M data sets, where M is a positive integer each of which includes at least one learning data in a learning mode and ii) extracting N heterogeneous object feature data from at least one object data to be processed, in an execution mode; a neural network including neurons and having N input layers and an output layer for learning a relation between each of the N learning feature data and a teacher data corresponding to the N learning feature data and a relation among the N learning feature data of each data set based on the N learning feature data of each set from said feature extracting means and a corresponding teacher data supplied thereto, in the learning mode, and for outputting a target data based on the learning result in response to input of the N object feature data from said feature extracting means in the execution mode, particular ones of said neurons being operatively coupled to the N input layers, respectively, such that the particular ones of said neurons can learn the relation among the N learning feature data; and, learning control means for substantially simultaneously supplying each set of the N learning feature data and a corresponding teacher data to said neural network and for controlling said neural network to learn the relations in the learning mode.
Claim 2
Dependent Claim
247 words
A neural network system comprising: feature extracting means for i) extracting N heterogeneous learning feature data, where N is an integer greater than one, from each of M data sets, where M is a positive integer each of which includes at least one learning data in a learning mode and ii) extracting N heterogeneous object feature data from at least one object data to be processed, in an execution mode; a neural network including neurons and having N input layers and an output layer for learning a relation between each of the N learning feature data and a teacher data corresponding to the N learning feature data and a relation among the N learning feature data of each data set based on the N learning feature data of each set from said feature extracting means and a corresponding teacher data supplied thereto, in the learning mode, and for outputting a target data based on the learning result in response to input of the N object feature data from said feature extracting means in the execution mode, particular ones of said neurons being operatively coupled to the N input layers, respectively, such that the particular ones of said neurons can learn the relation among the N learning feature data; and, learning control means for substantially simultaneously supplying each set of the N learning feature data and a corresponding teacher data to said neural network and for controlling said neural network to learn the relations in the learning mode.
Claim 3
Dependent Claim
186 words
A neural network system comprising: a neural network, including N input layers where N is an integer greater than one and an output layer to which input data inputted to the N input layers are propagated, for receiving a learning data and a teacher data for learning in a learning mode and for outputting a target data based on the learning result in response to input of heterogeneous object data in an execution mode, the learning data comprising N heterogeneous learning patterns and the object data comprises N heterogeneous object patterns; supply means for simultaneously and respectively supplying to the N input layers of said neural network the learning data including the N heterogeneous learning patterns and the teacher data in the learning mode, and the object data including the N heterogeneous object patterns in the execution mode; and, learning control means for controlling the learning of said neural network while adding to weights of synapse connections between neurons in said neural network during the learning using noise data, having an intensity which is determined based on the number of prior-learned cycles, in the learning mode.
Claim 4
Dependent Claim
186 words
A neural network system comprising: a neural network, including N input layers where N is an integer greater than one and an output layer to which input data inputted to the N input layers are propagated, for receiving a learning data and a teacher data for learning in a learning mode and for outputting a target data based on the learning result in response to input of heterogeneous object data in an execution mode, the learning data comprising N heterogeneous learning patterns and the object data comprises N heterogeneous object patterns; supply means for simultaneously and respectively supplying to the N input layers of said neural network the learning data including the N heterogeneous learning patterns and the teacher data in the learning mode, and the object data including the N heterogeneous object patterns in the execution mode; and, learning control means for controlling the learning of said neural network while adding to weights of synapse connections between neurons in said neural network during the learning using noise data, having an intensity which is determined based on the number of prior-learned cycles, in the learning mode.
Claim 5
Dependent Claim
40 words
A neural network system according to claim 2, wherein said learning control means includes means for adding the noise data having an intensity based on a cooling schedule, said intensity decreasing in value as the number of prior-learned cycles increases.
Claim 6
Dependent Claim
99 words
A neural network system according to claim 2, wherein said learning control means includes: a noise array table storing a first noise intensity predetermined for each synapse connection of each neuron of said neural network; a noise cycle number table storing a second noise intensity determined based on the number of previously learned cycles; and means for generating and adding to the weight of each synapse connection, the noise data represented by a product of the first noise intensity of the noise array table and the second noise intensity of said noise cycle number table for the synapse connection.
Claim 7
Dependent Claim
125 words
A neural network system according to claim 2, wherein said neural network has two input layers and an output layer to which two data inputted to the two input layers are propagated respectively, and said supply means includes means for supplying the learning data and the teacher data to the two input layers at different timings, respectively, in the learning mode and the object data to one of the two input layers to which the learning data is supplied in the execution mode; said learning control means includes means for controlling the learning of said neural network, based on a difference between outputs respectively corresponding to the learning data and the teacher data from the output layer of said neural network in the learning mode.
Claim 8
Dependent Claim
92 words
A neural network system according to claim 2, wherein said learning control means comprises: means for setting an initial value of a coefficient of each synapse connection in said neural network; means for changing each synapse connection coefficient in accordance with the teacher data while in the learning mode; and repeating means for causing said changing means to operate until a difference between output data of said neural network and the teacher data is less than a predetermined value while feeding the difference back to the input layer of said neural network.
Claim 9
Dependent Claim
91 words
A neural network system according to claim 2, wherein said learning control means includes means for changing each synapse connection coefficient in accordance with a predetermined cost function being a sum of a function of the connection coefficient reflecting selection tendency of the connection coefficient providing a place of storage and a function providing the sum, for each input combination, a square error between the output at the output layer obtained from the sum of the products of the coefficients of synapse connection and each neuron condition and the teacher data.
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) | 1996-10-18 | 1997-04-19 | 1997-10-19 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2000-10-18 | 2001-04-19 | 2001-10-19 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2004-10-18 | 2005-04-19 | 2005-10-19 | $7700.00 | Window Closed / Lapsed |