Method of configuring a neural network and a diagnosis/control system using the neural network
First Claim
Patent Images
1. A neural network system, comprising:
- sensor means including a plurality of sensors for sensing a plurality of states of an object;
network means including a neural network for outputting data for the object based on the sensed states of the object;
means for determining a sum of a plurality of input data constituting a data pattern set which includes teaching data, said means for determining a sum of a plurality of input data determines a plurality of sums of said input data respectively for a plurality of data pattern sets;
means for arranging the plurality of sums of said input data in descending order to obtain an arrangement of respective teaching data;
configuring means for configuring a combination of neurons constituting said neural network by forming a logical operation circuit for outputs of neurons in accordance with the plurality of arranged teaching data.
1 Assignment
0 Petitions
Accused Products
Abstract
The number of hidden layers is no larger than 2 and a sum is determined for each case. A relationship between the sum of the inputs and teaching data for each case is expressed in a table in a descending order of the sum of inputs for each output of an output layer, and the teacher data for the maximum sum and the number of times of change in the teacher data are considered. A configuration (the number of hidden layers and the number of neurons thereof) is determined based on those data, and the coupling weights can be analytically calculated by using the table. Where the number of times of change of the teacher data is odd, some inputs do not route a second hidden layer.
29 Citations
23 Claims
-
1. A neural network system, comprising:
-
sensor means including a plurality of sensors for sensing a plurality of states of an object; network means including a neural network for outputting data for the object based on the sensed states of the object; means for determining a sum of a plurality of input data constituting a data pattern set which includes teaching data, said means for determining a sum of a plurality of input data determines a plurality of sums of said input data respectively for a plurality of data pattern sets; means for arranging the plurality of sums of said input data in descending order to obtain an arrangement of respective teaching data; configuring means for configuring a combination of neurons constituting said neural network by forming a logical operation circuit for outputs of neurons in accordance with the plurality of arranged teaching data. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12)
-
-
13. A method for configuring a neural network, said method being executed by a computer system and comprising the steps of:
-
determining a sum of a plurality of input data constituting a data pattern set which includes teaching data of binary value data, a plurality of sums of said input data being determined respectively for a plurality of data pattern sets; arranging the plurality of sums of said input data in descending order to obtain an arrangement of respective teaching data; and determining a configuration of a combination of neurons constituting said neural network in accordance with a number of transitions between binary values in a sequence of said plurality of teaching data arranged in descending order. - View Dependent Claims (14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A method for configuring a neural network, said method being executed by a computer system and comprising the steps of:
-
determining a sum of learning data for each of a plurality of learning cases; rearranging teaching data for the learning cases in descending order, all teaching data being binary-value data; determining a number N of transitions between binary-values of the teaching data; providing the number of neurons which is the same as the number N of transitions to a first hidden layer; providing (N-1)/2 neurons where N is odd and N/2 neurons where N is even to a second hidden layer; connecting an input layer to the neurons of said first hidden layer to continuously supply data "1" to the neurons of said first hidden layer; connecting outputs of two neurons of said first hidden layer to input of each neuron of said second hidden layer to continuously supply data "1" to each neuron of said second hidden layer; connecting an output of each neuron of said second hidden layer to an output layer where N is even and connecting the output of each neuron of said second hidden layer and the output of the remaining neuron of said first hidden layer where N is odd so that data "1" is continuously supplied to said output layer.
-
-
23. A diagnosis system using a neural network system, comprising:
-
sensor means including a plurality of sensors for sensing a plurality of states of an object; network means including a neural network for outputting. data for the object based on the sensed state of the object; means for determining a sum of a plurality of input data constituting a data pattern set which is accompanied with teaching data, said means for determining a sum of a plurality of input data determines a plurality of sums of input data respectively for a plurality of data pattern sets; means for arranging the plurality of sums of input data in descending order to obtain a sequence of a plurality of respective teaching data, the plurality of sums of input data arranged in descending order and the plurality of respective teaching data arranged in descending order being stored in table form; means for inputting a plurality of newly input diagnosis data constituting a data pattern set having unknown teaching data; means for determining a sum of the plurality of newly input diagnosis data; means for determining nearest sums along the plurality of sums of input data in said table for the sum of said newly input diagnosis data; and means for determining teaching data for the plurality of newly input diagnosis data based on teaching data corresponding to the nearest sums in said table.
-
Specification