Patent EP0488150A2 successfully refreshed from Google Patents.
EP0488150A2
Expired / Lapsed
EP
⏱️ 0.0 years remaining
Neural network system adapted for non-linear processing
Assignee: Hitachi Ltd
•
Filing: 1991-11-26
•
Granted: —
•
Expires: 2011-11-26
Abstract
Abstract A neural network system includes a qualitative evaluation section (2), a neural network section (5), a quantifying section (3) and a display section (32). The qualitative evaluation section (2) qualitatively analyzes an unknown data supplied thereto, and normalizes the result of analysis within a predetermined range. The neural network section (5) having a neural network (6) with plural neurons computes the network output data from the normalized unknown data produced by the qualitative evaluation section (2). Each neuron is connected to plural of neurons through synapses, each of which is assigned with a weight coefficient. Each neuron is adapted to output an output function value assigned thereto associated with the total sum of the products of the output from the neurons connected thereto and the synapse weight coefficient. The quantifying section (3) quantifies the network output data to produce desired data. The desired data thus produced is displayed on the display section. <IMAGE>
Inventors
- 👤 Haruki Inoue
- 👤 Kenichi Nakamura
- 👤 Keiji Oshima
- 👤 Masakazu Yahiro
- 👤 Minoru Koide
- 👤 Noboru Abe
Bibliographic Details
Application Number
—
Jurisdiction
EP
Data Source
google_patents
Last Synchronized
2026-10-30 04:01
CPC Classifications
G; G06; G06J; G06J3/00; G06N; G06N3/00; G06N3/02; G06N3/08; G06N3/04; G06N3/042; G06N3/043; G06N3/0495; G06N3/0499; G06N3/082; G06N3/09; G06N7/00; G06N7/02; G06N7/04; G06N7/046; Y; Y10; Y10S; Y10S706/00; Y10S706/90
Claims Analysis
Total Claims: 51
Claim 1
Independent Claim
87 words
A neural network system comprising: normalizing means (11) for normalizing input data within a predetermined range; and neural network means (6, 8) including a neural network having a plurality of neurons, for computing output data from the normalized data from the normalizing means, wherein each neuron is connected to neurons through synapses, each of which has a weight, each neuron outputting a value of an output function assigned thereto for a total sum of product values of output values from the connected neurons and corresponding synapse weight.
Claim 2
Dependent Claim
72 words
A neural network system according to Claim 1, further comprising correcting means (7) for correcting the weight of each synapse of said neural network means (6, 8) in accordance with a difference between teacher data and the output data from said neural network means (6, 8) when learning data is supplied as the input data to said normalizing means (11), and thereby causing said neural network means to learn the learning data.
Claim 3
Dependent Claim
94 words
A neural network system according to Claim 1, further comprising table storage means (10) for storing a product value table (10-1) which stores product values and a function value table (10-2) which stores output function values, and wherein said neural network means (6) further includes means (61, 62, 63) for referring to said product value table to determine a product value for each of the connected neuron, determining the total sum of the product values, and referring to said function value table in accordance with the total sum to determine the output function value.
Claim 4
Dependent Claim
34 words
A neural network system according to Claim 1, further comprising qualitative analysis means (2) for qualitatively analyzing data supplied thereto and supplying a result of analysis as the input data to said normalizing means.
Claim 5
Dependent Claim
43 words
A neural network system according to Claim 1, further comprising: display means (32) for displaying data inputted thereto; and qualitative evaluation means for qualitatively analyzing the output data from said neural network means and outputting a result of analysis to said display means.
Claim 6
Dependent Claim
148 words
A neural network system comprising: table storage means (10) for storing a product value table (10-1) which stores product values and a function value table (10-2) which stores output function values; and neural network means (6, 8) including a neural network having a plurality of neurons, for computing output data for input data, each neuron being connected to neurons through synapses each having a weight, each neuron outputting a value of an output function for a total sum of product values of output values from the connected neurons and corresponding synapse weights, and each neuron referring to said product value table to determine a product value for each of the connected neurons, determining the total sum of the product values, and referring to said function value table to determine the output function value for the total sum, thereby to finally compute the output data from the input data.
Claim 7
Dependent Claim
74 words
A neural network system according to Claim 6, further comprising correcting means (7) for correcting the weight of each synapse of said neural network means (6, 8) in accordance with a difference between teach data and the output data from said neural network means (6, 8) and to supply thereto of learning data as the input data and causing said neural network means (6, 8) to learn the learning data in the learning mode.
Claim 8
Dependent Claim
53 words
A neural network system according to Claim 6, further comprising removing means (9) for removing computation on the synapse whose weight is smaller than a predetermined value and a neuron, synapse weights connected to which are smaller than the predetermined value, from computations for obtaining the output data based on a normalized data.
Claim 9
Dependent Claim
54 words
A neural network system according to Claim 1,3 or 6, further comprising removing means (9) for removing computation on a synapse whose weight is smaller than a predetermined value and a neuron, all synapses connected to which are smaller than the predetermined value, from computation for obtaining the output data from a normalized data.
Claim 10
Dependent Claim
54 words
A neural network system according to Claim 1 or 6, further comprising: means (10') for storing a qualitative causal network representing relation between input and output data; and network defining means (12) for connecting the neurons to each other through the synapses in accordance with the qualitative causal network to define said neural network.
Claim 11
Independent Claim
62 words
A method of constituting a neural network, comprising the steps of: storing a qualitative causal network in storage means representing qualitative causal relation between a plurality of input data and output data; arranging neurons in accordance with a plurality of input data and output data; and connecting the arranged neurons to constitute a neural network in accordance with the qualitative causal network.
Claim 12
Dependent Claim
54 words
A method according to Claim 11, wherein: said connecting step includes the steps of forming subnetworks by connecting neurons of a given layer to a neuron of a subsequent layer in accordance with the qualitative causal network; and constituting the neural network by connecting the formed subnetworks in accordance with the qualitative causal network.
Claim 13
Dependent Claim
99 words
A neural network system comprising: means (10') for storing a causal network representing qualitative causal relation between input data and output data; network defining means (12) for connecting a plurality of neurons through synapses defining a neural network in accordance with the causal network; and neural network means (6, 8) including said neural network in which each neuron is connected to neurons through synapses, each synapse being given a weight, each neuron outputting a value of an output function attached thereto associated with a total sum of products of output values from the connected neurons and the synapse weights.
Claim 14
Dependent Claim
70 words
A neural network system according to Claim 13, further comprising: correcting means (7) for correcting the weight of each synapse of said neural network in accordance with difference between output data from said neural network means and a teacher data in response to supply of learning data to said neural network means as input data, thereby causing said neural network means to learn the learning data in a learning mode.
Claim 15
Dependent Claim
100 words
A neural network system according to Claim 13, further comprising table storage means (10) for storing a product value table (10-1) which stores product values and a function value table (10-2) which stores output function values, and wherein said neural network means (6, 8) further includes means for referring to said product value table to determine a product value for an output function value of each of the connected neurons determining the total sum of the product values, and referring to said function value table in accordance with the total sum to determine an output function value of each neuron.
Claim 16
Dependent Claim
68 words
A neural network system of Claim 3, 6 or 15, wherein the neurons of said neural network are layered, and said neural network means (6, 8) further includes means (61, 62, 63) for causing said neural network to synchronously perform processing for referring to said product value table, processing for determining the total sum, and processing for referring to said function value table, in units of layered neurons.
Claim 17
Dependent Claim
36 words
A neural network system of Claim 6 or 13, further comprising normalizing means (11) for normalizing data supplied thereto within a predetermined range and supplying the neural network with the normalized data as the input data.
Claim 18
Dependent Claim
43 words
A neural network system of claim 6 or 13, further comprising qualitative evaluation means (2) for qualitatively analyzing data supplied thereto, normalizing the result of analysis within a predetermined range, and supplying the normalizing data to said neural network as the input data.
Claim 19
Dependent Claim
44 words
A neural network system at Claim 6 or 13, further comprising: display means (32) for displaying data inputted thereto; and qualitative evaluation means for qualitatively analyzing the output data from said neural network means and outputting a result of analysis to said display means.
Claim 20
Dependent Claim
35 words
A neural network system of Claim 2, 7 or 14, further comprising qualitative evaluation means for qualitatively analyzing data inputted thereto and supplying the result of analysis to said correcting means as the learning data.
Claim 21
Dependent Claim
32 words
A neural network system of Claim 1, 6 or 13, further comprising quantifying means (3) for quantitatively analyzing the output data from said neural network means and outputting a result of analysis.
Claim 22
Dependent Claim
154 words
In a neural network system, a method of determining an optimal solution, comprising the steps of: qualitatively normalizing learning data within a predetermined range, generating output data from the qualitatively normalized learning data by a neural network including a plurality of neurons, each neuron being connected to neurons through synapses, each synapse being given a weight, each neuron outputting a value of an output function given to a particular neuron associated with a total sum of the product values of output function values of the connected neurons and the synapse weights; correcting the weight of each synapse in said neural network in accordance with difference between an output of said neural network and teacher data and causing said neural network to learn the learning data; qualitatively normalizing unknown data within the predetermined range; and generating desired data from the qualitatively normalized unknown data by the neural network in accordance with the result of learning.
Claim 23
Dependent Claim
28 words
A method according to Claim 22, further comprising the step of converting the desired data from the neural network into a quantative value and outputting the quantitative value.
Claim 24
Dependent Claim
26 words
A method according to Claim 22 further comprising the step of qualitatively analyzing the desired data from the neural network and outputting the result of analysis.
Claim 25
Dependent Claim
50 words
A method according to Claim 22, wherein each of the steps of generating output data and generating desired data includes the step of invalidating the synapses whose weights are smaller than a predetermined value, and a neuron, all the synapses weights connected to which are smaller than the predetermined value.
Claim 26
Dependent Claim
118 words
A method according to Claim 22, wherein each of the steps of generating output data and desired data includes the steps of: (a) determining product values of weights of synapse and output function values from neurons connected to a given neuron through synapses, with reference to a product value table for storing product values, in a process for generating selected one of the output data and the desired data; (b) determining a total sum of the product values; (c) determining an output of the neuron with reference to a function value table for storing output function values in accordance with the total sum; and (d) executing steps (a) to (b) for all the neurons in said neural network.
Claim 27
Dependent Claim
119 words
A neural network system comprising: first qualitative evaluation (2) means for qualitatively analyzing unknown data supplied thereto and normalizing the result of analysis within a predetermined range; computing means (5) including a neural network having a plurality of neurons, for computing desired data from the normalized unknown data from said first qualitative evaluation means, each neuron being connected to neurons through synapses, each synapse being given a weight, each neuron outputting a value of an output function given to the neuron in accordance with a total sum of product values of output function values from the connected neurons and the synapse weights; and second qualitative evaluation means for qualitatively analyzing the desired data and outputting the result of analysis.
Claim 28
Dependent Claim
57 words
A neural network system according to Claim 27, wherein said computing means further comprises removing means (9) for removing computation on the synapses whose weights are smaller than a predetermined value, and a neuron, the synapses connected to which are smaller than the predetermined value, from the computation of the desired data based on the unknown data.
Claim 29
Dependent Claim
98 words
A neural network system according to Claim 27, wherein said computing means (5) further comprises: table storage means (10') for storing a product value table (10-1) which stores product values and a function value table (10-2) which stores output function values; and means (61, 62, 63) for determining a product value for each output function value of neurons connected to said each neuron with reference to said product value table, determining a total sum of the product values, and determining an output function value of neurons with reference to the function value table according to the total sum.
Claim 30
Dependent Claim
61 words
A neural network system according to Claim 29, wherein the neurons of said neural network are layered, and said computing means (5) further includes means for causing said neural network to synchronously perform processing for referring to said product value table, processing for determining the total sum, and processing for referring to said function value table, in units of layered neurons.
Claim 31
Dependent Claim
56 words
A neural network system according to Claim 29, wherein said computing means (5) further includes removing means (9) for removing computation on the synapse whose weight is smaller than a predetermined value and a neuron' synapse weights connected thereto whose are smaller than the predetermined value, from computation of the desired data from the unknown data.
Claim 32
Dependent Claim
55 words
A neural network system according to Claim 27, wherein said computing means (5) further includes: means (10') for storing a causal network representing qualitative causal relation between the unknown data and the desired data; and network defining means (12) for defining said neural network by connecting neurons through synapses in accordance with the causal network.
Claim 33
Dependent Claim
111 words
A neural network system comprising: first evaluation means (2) for qualitatively analyzing the unknown data supplied thereto and normalizing the result of analysis within a predetermined range; computing means (5) including a neural network having a plurality of neurons, for computing desired data from the normalized unknown data outputted from the first evaluation means, each neuron being connected through synapses to neurons, each synapse being given a weight, each neuron outputting an output function value in accordance with a total sum of products of output function values from the connected neurons and synapse weights; and second evaluation means (3) for quantitatively analyzing the desired data and outputting the result of analysis.
Claim 34
Dependent Claim
57 words
A neural network system according to Claim 33, wherein said computing means (5) further includes removing means (9) for removing computation on the synapse whose weight is smaller than a predetermined value, and a neuron, weights of synapses connected to which are smaller than the predetermined value, from computation of the desired data from the unknown data.
Claim 35
Dependent Claim
118 words
A neural network system according to Claim 33, wherein said computing means (5) further includes table storage means (10') for storing a product value table which stores product values and a function value table which stores output function values; and means (61, 62, 63) for determining the product for each output function value of neurons referring to the product value table in response to the supply of the output function values of the neurons connected to said each neuron, determining the total sum of the product values thus obtained, and determining an output function value with reference to the function value table associated with the total sum thereby to determine the output function value of the particular neurons.
Claim 36
Dependent Claim
60 words
A neural network system according to Claim 35, wherein the neurons of the neural network are tiered, and said computing means (5) further includes means (61, 62, 63) for determining the output function value of the neurons in a given layer with reference to the function value table subsequent to the product value table in synchronism with the particular neurons.
Claim 37
Dependent Claim
64 words
A neural network system according to Claim 33 wherein said computing means further includes removing means (9) for removing synapses of the neural network means whose weight coefficients are smaller than a predetermined value, and also a neuron and synapses connected thereto whose weight coefficients are all smaller than a predetermined value, from the computation of a desired data based on an unknown data.
Claim 38
Dependent Claim
52 words
A neural network system according to Claim 33, wherein said computing means (5) further includes means (10') for storing a network of qualitative causal relation between unknown and desired data, and network definition means (12) for defining the neural network by connecting neurons through synapses in accordance with the qualitative causal network.
Claim 39
Independent Claim
109 words
A method of fuzzy-reasoning a target data from an unknown data, using a neural network, said method comprising the steps of: learning the weight of each synapse for determining the neuron connection of a neural network from a first known data corresponding to a unknown data and a second known data providing a target data corresponding to the first known data, converting an unknown data into a qualitative evaluation by a fuzzy membership function, inputting the converted evaluation to the neural network, determining a value for each type of evaluation by the neural network, and determining a target data by defuzzy conversion of a set of evaluations thus determined.
Claim 40
Dependent Claim
109 words
A neural network system comprising: first fuzzy means (2) for fuzzy conversion of an unknown data supplied thereto, neural network means (5) including a neural network having a plurality of neurons for computing the network output data from the unknown data fuzzy-converted from the first fuzzy means, each neuron being connected through synapses to neurons, each synapse being given a weight coefficient, each neuron being adapted to output an output function value assigned thereto associated with the total sum of the products of the output from the neurons connected and the synapse weight coefficient, and defuzzy means (3) for defuzzy-converting the network output data and outputting a desired data.
Claim 41
Dependent Claim
64 words
A neural network system according to Claim 40, wherein said neural network means further includes removing means (9) for removing synapses in the neural network whose weight coefficients are smaller than a predetermined value, and also a neuron and synapses connected thereto whose weight coefficients are all smaller than predetermined value, from the computation of the network output data based on an unknown data.
Claim 42
Dependent Claim
119 words
A neural network system according to Claim 40, wherein said neural network means further includes table storage means (10) for storing a product value table (10-1) which stores product values and a function value table (10-2) which stores output function values, and means (61, 62, 63) for determining the product value for each output function value of neurons with reference to the product value table in response to the supply thereto of the output function value of neurons connected to each neuron, determining the total sum of the product values, and determining an output function value with reference to the function value table associated with the total sum thereby to determine the output function value of the particular neurons.
Claim 43
Dependent Claim
58 words
A neural network system according to Claim 42, wherein the neurons of said neural network are layered, and said neural network means (5) further includes means for determining the output function value of the neurons in a given layer with reference to the function value table subsequent to the product value table in synchronism with the particular neurons.
Claim 44
Dependent Claim
66 words
A neural network system according to Claim 40, wherein said neural network means (5) further includes removing means (9) for removing the synapses of the neural network whose weight coefficients are smaller than a predetermined value, and also a neuron and synapses connected thereto whose weight coefficients are all smaller than a predetermined value, from the computation of a desired data based on an unknown data.
Claim 45
Dependent Claim
53 words
A neural network system according to Claim 40, wherein said neural network means (5) further includes means (10') for storing a network of qualitative causal relation between unknown and desired data, and network definition means (12) for defining the neural network by connecting neurons through synapses in accordance with the qualitative causal network.
Claim 46
Dependent Claim
87 words
A process controller comprising: input means (31, 39-1, 41) for supplying values corresponding to values of elements which substantially influence a process for every predetermined period of time; neural network inferring means (36) for inferring a prediction amount in accordance with the corresponding values from said input means; coverting means (39-2, 32) for converting the prediction amount from said neural network inferring means into an optimal operation amount; and control means (42) for controlling the process in accordance with the optimal operation amount from said converting means.
Claim 47
Dependent Claim
62 words
A process controller according to Claim 46, wherein said neural network inferring means (36) comprises: normalizing (11) means for normalizing each of the element values in an predetermined range; and inferring means (6) having a neural network including a plurality of neurons, for inferring the prediction amount in accordance with the normalized element values from said normalizing means by said neural network.
Claim 48
Dependent Claim
134 words
A process controller according to Claim 46, wherein said neural network inferring means (36) comprises: a product value table (10-1) for storing product values; a function value table (10-2) for storing output function values; and inferring means (6) having a neural network including a plurality of neurons, each neuron being connected to neurons via synapses, each synapse being assigned with a weight, for referring to said product value table in accordance with each of the connected neurons and a corresponding one of the synapses for each neuron to determine a product value of each connected neuron and the corresponding synapse, determining a total sum of the determined product values for each neuron, and referring to said function value table in accordance with the total sum to determine an output function value for each neuron.
Claim 49
Dependent Claim
52 words
A process controller according to Claim 46, wherein said neural network inferring means (36) includes a neural network (6) with a pluralityof neurons, and the neurons are arranged within said neural network in accordance with a qualitative causal network which represents causal relation between a plurality of input information and output information.
Claim 50
Dependent Claim
26 words
A process controller according to Claim 46, wherein said input means (41, 31, 39-1) further includes qualitatively evaluating each element value to obtain the corresponding values.
Claim 51
Dependent Claim
26 words
A process controller according to Claim 46, wherein said converting means (39-2, 32) further includes quantitatively evaluating the prediction amount to obtain the optimal operation amount.
References Cited (Backward Citations)
No citation records indexed.
USPTO Annuity & Maintenance Windows
Maintenance schedule is calculated for granted US utility patents.