JPH0424741A - Learning method for prewired neuro of rule part - Google Patents

Abstract

PURPOSE:To efficiently adjust the weight of the antecedent implementing part of a membership function by learning only the weight of the antecedent implementing part of the membership function after initializing a weight value for the antecedent implementing part of the membership function, a rule part and the consequent implementing part of a membership function. CONSTITUTION:An initial weight value to which preliminarily stored knowledge is introduced is set for each of an antecedent implementing part 10 of the membership function, a rule part 12 and a consequent implementing part 14 of the membership function. Then, the weight of the antecedent part 10 is learned by using a learning data and after that, the weight of the rule part is learned. The weight value can be initialized for the antecedent part 10 by using a random number or by using both the preliminarily stored knowledge and the random number. Thus, it is possible to efficiently learn to adjust the weight of the antecedent implementing part of the membership function.

Description

[Detailed Description of the Invention] [Summary] A control output corresponding to an input is generated according to a hierarchical network structure that realizes a fuzzy control rule, which is composed of an antecedent membership function implementation unit, a rule unit, and a consequent membership function implementation unit. Regarding the learning method of the rule part pre-wired neuro, we have developed an antecedent membership function realization part, a rule part and a consequent part membership function realization part, with the aim of efficiently learning to perform control suitable for the target system. After setting an initial weight value that introduces pre-existing knowledge for each of The configuration is such that the weights of the rule section are learned after the weight learning is completed. Note that the initial setting of the weight value of the antecedent membership function realization unit may be performed using random numbers, or may be performed using a combination of knowledge held in advance and random numbers.

[Industrial Application Field] The present invention relates to a learning method for a rule section prewired neuron that combines fuzzy control and a neural network.

In information processing using neurons, learning data consisting of a set of input patterns of the target system and desired output patterns for that input pattern is used in a hierarchical format.
It is presented to the physical network to learn and perform adaptive processing. In particular, a processing method called pack propagation method is attracting attention due to its high practicality in this learning process.

On the other hand, fuzzy theory was proposed by 2adeh in 1960, and is a theory that models the process of thinking and judgment based on ambiguity that humans make, such as whether the temperature is "low" or "high." This fuzzy theory is mathematically based on fuzzy set theory, which introduces membership functions to represent ambiguity.

As an application of fuzzy theory, in 1974 Maqmda
Various types of fuzzy control have been used since ni was first used to control steam engines.

Since the system that performs this fuzzy control handles uncertain things, it is necessary to have a configuration that can adaptively respond to adjustments and changes in the control algorithm.

[Prior Art] Generally, the following procedure is taken to construct a fuzzy control system.

■Obtain ambiguous control rules such as ``If the temperature is high, reduce the fire'' held by experienced operators.

■The meaning of the words (propositions) in the control rule is quantified in the form of a membership function, and the control rule is described in a rule expression of the type ``rlF-THEN part''.

■Conduct inspections through simulations and on-site tests.

■Evaluate inspection results and improve membership functions and rules.

To explain in more detail, the fuzzy control rules are, for example, lit x, is big and xlis sm
It is written as all then7, is bigJ. Here, the IF section is called an antecedent section, and is a section that describes conditions regarding control state quantities (control inputs) such as temperature. Further, the THEN section is called a consequent section, and is a section that describes the conditions regarding the control operation amount (control output) of the operating terminal, etc.

Fuzzy control manages multiple fuzzy control rules prepared as the control logic of the target system, and also manages the meaning of ambiguous linguistic expressions such as "large" or "small" written in each fuzzy control rule. Quantify and manage it as a ship function.

The control procedure of the completed fuzzy control system is as follows.

■When a controlled state quantity such as temperature is input from the controlled object, the truth value is calculated according to the membership function of the antecedent part that has been set.

(2) Next, in accordance with the antecedent computation such as selecting the minimum value, a process is executed to determine the value to be applied to the consequent in each fuzzy control rule. For example, "xlis big
The truth value of J = 0.8 "If the truth value of the Process so that it is output as an applied value for the part.

(2) Next, a process is executed to determine an output application value from the input value (applied value) from the antecedent part to the membership function of the consequent part, which has the same amount of control operation.

For example, with two fuzzy control rules, [yl i
s b + gJ applied value = 0.5 "y, is bigJ
When the applied value=0.6 is obtained, the applied value=0.6 is selected and determined according to the consequent operation that selects the maximum value.

■Subsequently, the membership function of the control operation amount is reduced (or expanded) according to the determined applied value, and the center of gravity of the figure of the sum of the membership functions for the same control operation amount is calculated. The control operation amount, which is a fuzzy inference value, is calculated and output.

By the way, fuzzy control rules are planned to be generated according to the knowledge possessed by a skilled operator, and it is difficult to establish fuzzy control rules that can realize optimal control from the beginning. Therefore, it is necessary to improve the generated fuzzy control rules through trial and error while evaluating them through simulations and field tests.

Therefore, the inventors of the present application developed a hierarchical network structure that realizes fuzzy control rules by applying neurology.
That is, we have proposed an adaptive fuzzy control system using prewired neuro in the rule section (Patent Application No. 2-602).
(See "Hierarchical Network Configuration Data Processing Device and Data Processing System L," No. 56). That is, the rule part prewire neuro has a hierarchical network configuration that realizes a fuzzy control rule consisting of an antecedent part membership function realization part, a rule part, and a consequent part membership function.

Considering the automatic extraction of fuzzy control rules by the pre-wired neuron of the rule section, the following procedure may be taken.

■Set the membership function of the rule section prewire neuro.

■The rule section prewire neuro is presented with the input pattern of the target system and a set of desirable output patterns for that input pattern, and is made to learn.

■Extract the optimal fuzzy control rule for the target system by analyzing the weights of connections in the rule section.

[Problems to be Solved by the Invention] However, the learning performed for extracting fuzzy control rules for the rule section pre-wired neuron is
The learning method for learning, including the weights of the antecedent membership function part, rule part, and consequent part membership function realization part, has not been determined, and the performance of the roll part prewire neuron has not yet been determined. There is a need to develop efficient learning methods that can bring out the best in students.

The present invention has been made in view of the above situation, and an object of the present invention is to provide a learning method for a rule section pre-wired neuron that can efficiently perform learning for controlling a target system. .

[Means to solve the problem] FIG. 1 is a diagram explaining the principle of the present invention.

First, the present invention provides antecedent membership function realization unit 10.1
Or it has a hierarchical network structure consisting of a rule part 12 and a consequent part membership function realization part 14 which are made up of multiple hierarchies,
The rule section 12 is an antecedent membership function realization section 10 in the previous stage.
and/or with the consequent membership function realization unit 14 in the subsequent stage, without internally connecting all the units, some of them are internally combined according to the control rule, and the input control state quantity (Xl
.

Rule section pre-wire neuro 10 that outputs one or more control operation amounts (Y) corresponding to X2+...Xo)
Target 0.

In the present invention, as a learning method for such a rule part pre-wired neuron 100, [First process] Each of the antecedent part membership function realization part 10, the rule part 12, and the consequent part membership function realization part 14 Initialize the weight values for; [Second process: learn only the weights of the antecedent membership function realizing unit 10 using the training data prepared in advance; [Third process] Rule part 12 using learning data
A learning method is constructed that includes a processing process in which only the weights of are learned;

Here, as the initial setting of the weight value of the first process, the knowledge (in the figure, "first" is Initialize weight values created based on "experimental knowledge").

In addition, as another initial setting of the weight value in the first process, a weight value created using random numbers is initially set in the antecedent membership function realization unit 10, and the rule unit 12 and the consequent membership function realization unit 10 are initialized with weight values created using random numbers. The unit 14 is initially set with weight values created based on knowledge held in advance.

Furthermore, the antecedent membership function realization unit 1 in the first process
Another method for setting the initial weight value of 0 may be to initialize weights that can be set using knowledge held in advance, and initialize other weights using random numbers.

On the other hand, at the start of learning in the second process, the learning flag provided for each connection of the antecedent membership function realization unit 10 is turned on, and the learning flag provided for each connection of the rule unit 12 and the consequent membership function realization unit 14 is turned on. The provided learning flag is turned off, and only the weight values of the connections having the learning flag in the on state are optimized through learning processing.

Furthermore, at the start of learning in the third process, a learning flag provided for each connection of the rule section 12 is turned on, and a learning flag provided for each connection of the antecedent membership function realization section 10 and the consequent membership function realization section 14 is turned on. The learning flag that has been set is turned off, and only the weight values of the connections that have the learning flag that is on are optimized through learning processing.

[Operations] According to the learning method of the rule section pre-wired neuron of the present invention having such a configuration, the following effects can be obtained.

First, the antecedent membership function realization unit of the rule unit prewire neuro is initialized with weight values based on prior knowledge, weight values using random numbers, or a combination of both. For the rule part and the consequent part membership function realization part, initial setting of weight values is performed based on knowledge held in advance.

Next, the learning process is started, the input cover turn for learning is presented to the rule unit pre-wire neuron, and the output cover turn from the rule unit pre-wire neuron is a learning ratio cover turn (the teacher For example, learning of weights by a pack propagation method is performed only for the antecedent membership function realization unit, in which the initially set weight values are corrected so as to substantially match the signal).

After the weight learning of the antecedent membership function realizing section is completed, next weight learning is performed on the rule section using the same backpropagation method.

In this way, by learning the weights of the antecedent membership function realization part of the rule part prewire neuron and then learning the weights of the rule part, pre-existing knowledge can be introduced and initial settings can be made. While effectively utilizing the weights of the rule part and the membership function realization part of the consequent part, it is possible to adjust the weight of the antecedent part membership function realization part, and also to adjust the weight of the rule part.

In addition, the initial setting of the weights of the antecedent membership function realization unit into which pre-existing knowledge is introduced facilitates learning and improves learning efficiency.

Furthermore, by initializing the weights of the antecedent membership function realization unit using random numbers, it is possible to maximize the neuro-based learning function.

[Embodiment] FIG. 2 is a learning operation flowchart showing a processing procedure as a first embodiment of the learning method for the rule section pre-wire neuro according to the present invention.

The rule section pre-wired neurons that are the targets of the learning method according to the procedure of FIG. 2 are, for example, the antecedent membership function realization section 10, the rule section 12, and the consequent membership function realization section 14 shown in FIG. (including the centroid calculation implementation unit 18) has a hierarchical network structure for realizing fuzzy control rules.

Here, the rule section pre-wire neuro means the rule section 12.
A network structure is created between the antecedent membership function realization unit 10 in the previous stage and/or the consequent membership function realization unit 14 in the subsequent stage, in which all units are not internally connected, but some are internally connected according to control rules. This is compared to a rule-based fully connected neuron that internally connects everything between hierarchical units.

In the learning process of FIG. 2, the following steps 81 to S6
Performs processing consisting of.

[Step S1 Weight Cab of the antecedent membership function realization unit 10.

Cbc is initialized based on the introduction of prior knowledge to realize a given membership function.

[Step S2: Initialize the weights Ccd and Cde of the co-rule unit 12 based on the introduction of pre-existing knowledge so as to realize the given fuzzy rule.

[Step S3] The weight Cej of the consequent membership function realization unit 14 is initialized based on the introduction of knowledge held in advance so as to realize the given membership function.

[Step S4] The weight Cfg of the center of gravity calculation implementation unit 18 is initialized to realize center of gravity calculation.

[Step S5] A phase/group correspondence table indicating neuron groups for each phase in which learning processing is performed and a group/connection correspondence table indicating connection groups belonging to each neuron group are set. In the present invention, the weight learning of the antecedent membership function realizing unit 10 is performed in phase 1, and the weight learning of the rule unit 12 is performed in phase 2, so that the neuron groups belonging to the learning target part of phase 1.2 Create a phase/group correspondence table for

[Step S6] Start the learning scheduler and perform learning processing using the pack propagation method. This learning process is phase 1
In phase 2, weight learning is performed only for the antecedent membership function realization unit 10, and in phase 2, weight learning is performed only for the rule unit 12.

Hereinafter, steps 81 to 81 shown in the learning operation flow of FIG.
Each of S6 will be explained in detail.

[Configuration of Rule Section Prewire Neuroj The rule section prewire neuron to which the present invention is applied has a hierarchical network structure capable of realizing the fuzzy control rules shown in FIG. 3, for example.

In FIG. 3, each node shown in each part represents one basic unit, and two types of basic units are used: a sigmoid function unit shown by a black circle, and a linear function unit shown by a white circle.

FIG. 4 shows the configuration of the basic unit shown in FIG. 3.

In FIG. 4, the basic unit 20 constitutes a multi-power single output system, and includes a multiplication section 22 that multiplies a plurality of inputs by the weight value of each internal connection, and a multiplication section 22 that adds all the multiplication results of the multiplication section 22. It is composed of an adding section 24 and a threshold processing section 26 that performs threshold processing on the output of the adding section 24.

Here, if the first layer is the h layer and the second layer is the i layer, the multiplier 22
The addition unit 24 executes the calculation of equation (1) below, and the threshold processing unit 26 executes the calculation of equation (2) below.

X p l=ΣyphW+h
(1) yp, = 1/ (1+ e-Lx"-")
(2) However, h: unit number of the h layer i: unit number of the i layer p: pattern number of the input signal θl: threshold W of the first unit of the i layer + h: h: weight value x of the internal connection between the i layers, : Sum of products yph of inputs from each unit of the h layer to the 1st unit of the i layer = Output ypl from the 6th unit of the h layer for the input signal of the pth pattern: 1 of the i layer for the input signal of the pth pattern Output from the number unit Here, the above equation (2) is known as a sigmoid function, and is the threshold value processing of the basic unit indicated by the black circle shown in FIG.

On the other hand, a linear function unit in which y=Xp is set in the threshold value processing section 26 becomes a linear function unit indicated by a white circle in FIG. However, the input section, which is the first unit of the antecedent membership function realization section 10, does not necessarily have to be a linear function unit. The hardware configuration of the basic unit shown in FIG. 4 is shown in Japanese Patent Application No. 63-216865 titled ``Network Configuration Data Processing Apparatus.''

First, the antecedent membership function realization unit 10 takes each input variable Xl+X2+...Xn as input, and inputs the input variable Xn
Output the grade value of the membership function of ~Xn. In the figure, each membership function is a one-variable function for simplicity, but it may be a multi-variable function. In other words, it may be a membership function XI &X2 (SA) with variables Xn and X2 as inputs.

The rule section 12 includes an antecedent section rule section 12-1 and a consequent section rule section 12-2. The antecedent part rule part 12-1 is
Membership function X + (S
The grade values of A) to Xn(LA) are input, and the grade values of each rule are output.

Further, the consequent part rule part 12-2 contains the grade value L of the rule.
It inputs H3-1 to LH8-n and outputs grade values of membership functions y(SA) to y(LA) of output variables.

Further, the consequent membership function realization unit 14 (including the centroid calculation realization unit 16) calculates the membership function y (
Input the grade values of SA) to y (LA) and output the value of the output variable Y.

Here, the antecedent part rule part 12-1 and the consequent part rule part 12
Since both the input and output of the rule unit 12 including -2 are the grade values of the membership function, these can be cascaded and multi-stage inference is possible.

Further, the neurons connecting each module 10.12-1.12-2.14 can be shared. For example, the output neuron of the antecedent membership function implementation unit 10 and the input neuron of the antecedent rule unit 12-1 may actually be the same.

Further, in FIG. 3, each unit is described as one neuron, but if the function of the unit can be specified, it may be used as a gate circuit, an arithmetic unit, etc. that realizes the unit function instead of a neuron.

FIG. 5A shows a specific embodiment of the rule section pre-wired neuron of the present invention, taking as an example a case where two inputs, Xn and X2, are given to obtain a control output Y.
The configuration of each part will be explained in detail with reference to the drawings as follows.

The input section 16 is composed of separation units 28-1 and 28-2, which divides the inputs Xn and X2 into two and supplies them to the antecedent membership function realization section 10. The antecedent membership function realization unit 10 includes four sigmoid function units 30-1 to 30.
-4 and the same four linear function units 32-1 to 32
-3.

The rule section 12 includes, for example, sigmoid function units 34-1 to 34-8 that realize eight fuzzy rules, and the antecedent membership function realization section 10 connects each fuzzy rule according to the descriptions 1 to L8. It is being done. The outputs of the sigmoid function units) 34-1 to 34-8 are connected according to the fuzzy rules to two linear function units) 36-1 and 36-2 which perform consequent operations according to the fuzzy rules.

The consequent membership function realization section 14 is composed of, for example, nine linear function units) 38-1 to 38-9, and determines the membership function based on the two unit output values (grade values) output from the rule section 12. The ship function is reduced (or enlarged) and the function sum of the two reduced (or enlarged) membership functions is calculated for each identical manipulated variable position.

Furthermore, the center of gravity calculation implementation unit 18 is composed of a center of gravity determining element output device 40 and a center of gravity calculation device 42. The linear function units 40-1 and 40-2 of the center-of-gravity determining element output device 40 calculate the weighting of the unit output from the consequent membership function realization unit 14 by calculating the sum and calculating the two center-of-gravity calculation elements Y,... Y2 is calculated, and finally the center of gravity calculation device 42 calculates the two center of gravity calculation elements Y+
, Y2, and generates a control output Y as a fuzzy inference value.

FIG. 5B shows neuron groups and connection groups used to implement the learning method according to the present invention for the rule section pre-wired neuron of FIG. 5A.

In FIG. 5B, there are seven neuron groups, Ga-Gg, among which neuron group G in the input section
a and 2 of neuron group Gg of center of gravity calculation realization unit 18
As for the first one, it is excluded from the weight learning target by the pack propagation method because it merely performs separation and addition/synthesis. Therefore, there are five neuron groups that are subject to weight learning by the pack propagation method: Gb, Gc, Gd, Ge, and Gf.

On the other hand, there are six connection groups, Cab-ClH, representing the input connections of each two-neelon group, and the weight values of the input connections located in the previous stage of the neuron group to be learned are learned by the pack propagation method.

Regarding the learning operation flow shown in Fig. 2, which targets the rule section pre-wired neuron having the structure shown in Figs. 5A and 5B,
First, the initial setting of weight values for each part will be explained as follows.

[Initial setting of the antecedent membership function realizing unit] The initial setting of the weight value for the antecedent membership function realizing unit 10 is to set the weight of each connection included in the connection groups Cab and Cbc to the given membership function. Initial settings are made to achieve this.

FIG. 6 shows the antecedent membership function realization unit 10 of FIG. 5A.
The control human power X1 side is extracted and shown. In FIG. 6, the output value y output from the sigmoid function unit 30-1.30-2 is y=1/(1+e-"x-"
) (3).

In the initial setting of the weight value for the antecedent membership function realizing unit 10 of the present invention, based on the knowledge that is held in advance, for example, as the weight value W of the input connection of the sigmoid function unit 30-1, W=- 20 is set, and θ=-10 is set as the threshold value θ in the sigmoid function unit 30-1.

Further, the weight value W of the input connection of the other sigmoid function unit 30-2 is set to W=20, and at the same time, the threshold value θ of the sigmoid function unit 302 is set to θ=10. Note that the linear function unit 32-1. ．． 32-
The weight value of the input connection No. 2 is set to the initial weight value=1.

In this way, the sigmoid function unit 30-1. In, W〈
When the weight value and threshold value are initially set such that 0 θ<0, the unit output value Xn (SS) for the control output Xn shown in the characteristic curve 44 in FIG. It is possible to set a truth value calculation function for a membership function having a function shape such that (SS) becomes small.

Also, in the sigmoid function unit 30-2, W>0
By initially setting the weight value and threshold value such that θ〉0, the unit output value X1 (LA) for the control human power X1 shown in the characteristic curve 46 in FIG. It is possible to realize a function that calculates the truth value of a membership function whose function shape is such that .

Sigmoid function unit 30-1.3 shown in FIG.
The initial setting of the weight value and threshold for 0-2 is such that the desired antecedent membership function is also realized for the remaining sigmoid function unit 30-3.30=4 side in the antecedent membership function realization unit 10. Initial settings are made based on prior knowledge.

[Initial Setting of Rule Section] The initial setting of the weight value for the rule section 1-2 in FIG. 5A is to set the weight value of each connection belonging to the connection groups Ccd and Cde shown in FIG. 5B to realize the given fuzzy rule. Introduce the knowledge you already have and make the initial settings.

FIG. 8 is an explanatory diagram showing a specific example of initial setting of weight values for the rule section 12 together with fuzzy rules.

In FIG. 8, eight sigmoid function units 34-
1 to 34-8 correspond to fuzzy rules 1 to L8 shown on the right side, respectively. For each of the sigmoid function units 34-1 to 34-8, fuzzy rules are calculated from the linear function units 32-1 to 32-4 of the antecedent membership function realization section 10 located at the previous stage.
Input connections are made in accordance with ~L8. Furthermore, fuzzy rules are applied to the two linear function units 36-1 and 36-2 provided following the eight sigmoid function units 34-1 to 34-8, and the consequent parts of 1 to L8 (T HE
Connections are made in accordance with N).

Here, we will explain the function shape of the fuzzy logic operation based on the weight values initially set for the input connections of the sigmoid function units l-34-1 to 34-8 in the rule part J-2 of FIG. 8 and the threshold values set for the units themselves. Then it becomes as follows.

FIG. 9 shows one sigmoid function unit 34 of the rule section 12, in which the truth value X+ is obtained from the antecedent membership function realization section 10 located at the previous stage.

X2 is input, and output value y is generated by product-sum calculation and threshold processing.

The output value y output from the sigmoid function unit 34 in FIG. 9 is: S = Wx+ 'X + +WXl' X 2-θ
(4) y=1/ (1+e'-s)
It is calculated as (5). The unit output value y given by the above equation (5) is the weight value W of the input connection,
,.

By arbitrarily setting WX2 and threshold value θ, an appropriate function shape can be realized.

Specifically, the sigmoid function units 34-1 to 34-1 in FIG.
The numerical value shown in the input connection of 34-8 is set as the initial weight value, and the sigmoid function units 34-1 to 34-8
The numerical value shown below indicates the threshold value θ.

For example, in the sigmoid function unit 34-1 that realizes the fuzzy rule L1, w, , = 12, Wx2 = 1
2, θ=6 is set, and by setting these initial values, the sigmoid function unit 34-1 that realizes the fuzzy rule L1 performs threshold processing of the function shape shown by the three-dimensional coordinates in FIG. .

Hereinafter, similarly, sigmoid function units 34-2 to 34-
By initializing the weight values and threshold values for each of the 8 fuzzy rules 2 to L8, the function shapes shown in FIGS. 11 to 17 can be obtained.

Note that the linear function unit 36 provided at the output stage in FIG.
The weight value of the connection for -1.36-2 is initially set to 1, and the threshold processing is a linear function, and θ
=0 is set.

[Initial settings of the consequent membership function realizing unit] The initial setting of the weight values for the consequent membership function realizing unit 14 in FIG. 5A is as follows: The weight values of the connections belonging to the connection group Cel in FIG. 5B are preset. Initial settings are made to realize the given membership function based on the knowledge of the given membership function.

FIG. 18 is an explanatory diagram specifically showing the initial setting of weight values for the consequent membership function realization unit 14.

In FIG. 18, the consequent membership function realization unit 14
Nine linear function units 38-1 to 38- provided in
9, a connection input is performed from the linear function unit 36-1, 36-2 of the rule section 12 at the previous stage as shown in the figure. These nine units are 38-1 to 38-9
correspond to points in the range of 0 to 1 on the training line at equal intervals.

The initial setting of the weight values of the input connections for the linear function units 38-1 to 38-9 is performed by the unit 36- of the rule section 12.
Regarding the weight value of the connection of the output y (SS) from 1, the horizontal axis is the control ratio Y1 and the vertical axis is the unit output value. Weight value, 1, 1. 1.0.75.0.5, 0.25. 0
．． 0. Initialize 0.

Further, regarding the weight values of the input connections from the unit output y (LA) of the rule section 12, the weight values are changed from 0 to 0 from the unit 38-1 side to the unit 38-9 side according to the characteristic curve 48. 0.0.25.0.5, 0.75. 1
．． 1. Initialize 1.

Furthermore, the threshold value θ of the linear function units 38-1 to 38-9 is θ
=0.

Due to the initial setting of the weight values in the consequent membership function realization section 14, the unit output values y(SS) and y(LA ) is reduced by multiplication by the initial weight value, added, and output to the next center of gravity calculation implementation unit 18.

In FIG. 18, a weight value of 1 or less is initially set so as to reduce the unit output, but it is also possible to set a weight value of 1 or more to enlarge it.

[Initial Setting of Center of Gravity Calculation Implementation Section] The initial settings of weight values for the center of gravity calculation implementation section 18 shown in FIG. 5A are initialized so as to realize center of gravity calculation for the connection group ClH of FIG. 5B.

FIG. 19 shows a specific example of initial setting of weight values for realizing the center of gravity calculation performed on the center of gravity determining element output device 40 of the center of gravity calculation implementation unit 18.

For the two linear function units 40-1 and 40-2 provided in the center-of-gravity determining element output device 4o, the nine linear function units 38-1 provided in the consequent membership function realization unit 14 of the preceding stage are ~38-9 are connected. Then, as shown in the figure, weight values for calculating two gravity center derived values Y, , Y2 used in the gravity center calculation are set for this input connection.

In fuzzy control, a method is generally adopted in which the control output Y, which is a fuzzy inference value, is calculated by finding the center of gravity of a figure consisting of the sum of functions of reduced membership functions for the same control output position according to the following formula: do.

Y= f grade (y) tdr / f grade
e(Y)dy (6) In order to realize this equation (6), the linear function unit 40-1 is converted from the linear function units 38-1 to 38-9 of the consequent membership function realization unit 14.
The minimum value Y- of the control output Y is used as the weight value of the connection for
Assign weight values that increase from 0 at equal intervals up to the maximum value Y=1 starting from ○. On the other hand, the linear function unit 40
-2 input connection, maximum value of control output Y = 1
From 0 to the minimum value Y=0, weight values that decrease from 0 to -1, for example, are assigned at the same equal intervals as above.

By weighting the input connections to the linear function units 40-1 and 402, the linear function unit 3 in the previous stage
When the unit output values of 8-1 to 38-9 are 01 to C9, the linear function unit 40-1 is as follows: Y1=IC+ +0.875 C2+0.75C3+0
．． 625 C4+0.5 C6+0.375 C6+(
Output 1,25c7+LI25c8+OC9(7),
Moreover, the linear function unit 40-2 has Y1=OCI −0
, 125C2-0, 25C30, 375C40, 5C6
0,625C6-0,75C,-0,875C8+IC
Outputs 9(8).

The centroid derived values y, y2 output from the linear function unit 40-1, 40-2 are given to the centroid calculation device 42, and Y=Y2/(Y2-Yl>(9)
using the control output Y, which is the center of gravity, that is, the fuzzy inference value
is calculated.

By substituting the above equations (7) and (8) into the above equation (9), it can be seen that Y in the above equation (9) matches Y in the above equation (6).

With the above processing, the initial setting of the weight value for the rule section prewire neuron shown in steps 81 to S4 in FIG. 2 is completed, and the process proceeds to step S5.

[Learning schedule setting] In step S5 of FIG. 2, following the completion of the initial setting of the weight values, a learning plan, that is, a learning schedule, is set for causing the rule section prewire neuro to perform learning processing. . The setting of this learning schedule is the same as the setting of the phase/group correspondence table shown in Fig. 20(a).
Perform the two settings of the group/connection correspondence table shown in b).

First, the phase-group correspondence table specifies the neuron groups to be learned in each phase as the learning phase progresses. In the present invention, after learning only the weights of the antecedent membership function realization unit 10 in phase 1, only the weights of the rule unit 12 are learned in phase 2, as shown in FIG. 20(a). In this way, neuron groups Gb and Gc belonging to the antecedent membership function realizing section 10 are set in phase 1, and neuron groups Gd and Ge belonging to the rule section 12 are set in phase 2. On the other hand, the group/connection correspondence table shows the correspondence between each two-knee long group Ga-Gg in the rule section pre-wire neuron and the input connection groups Cab and CIH as shown in Figure 20(b).
Set as shown.

When the settings of the phase/group correspondence table and the group/connection correspondence table are completed, the process proceeds to step S6 in Fig. 2, where the learning scheduler is activated and the learning data is actually given to the rule section prewire neuron to pack it. Learning of weight values is performed using the propagation method.

[Processing and Configuration of Learning Scheduler] FIG. 21 shows a flowchart of the learning process of the rule section pre-wired neuron of the present invention. This processing flow diagram is, for example, the 22nd
First, the device configuration for learning processing that can be realized by the device configuration shown in the figure will be explained.
In the figure, a learning instruction unit 50 and a learning unit 70 are provided for the rule section pre-wired neuron 100.

The learning instruction unit 50 has a learning scheduler 52,
For the learning scheduler 52, there is a learning signal storage unit 54 that stores learning signals consisting of pairs of control inputs and desired control outputs for the control inputs, and a phase group storage unit 54 that stores the phase group correspondence table shown in FIG. A correspondence tape 56, a group/connection correspondence table 58 storing a group/connection correspondence table, and a learning convergence determination section 60 are provided.

On the other hand, the learning unit 70 is provided with a learning instruction reading section 72, a learning flag setting section 74, and a weight changing section 76 that changes weight values using the pack propagation method.

Furthermore, a feature of the device configuration in FIG. 22 is that a learning adjuster 80 is provided in the connection connecting each unit in the rule section prewire neuron 100. The learning adjuster 80 has the configuration shown in FIG. Note that the weight change unit 76 is shown together with the learning adjuster 80. Learning regulator 8
0 is provided with a flag storage section 82, a weight change information reading section 84, and a weight change amount adjustment section 86.

On the other hand, the weight change unit 76 is provided with a weight calculation unit 88 that performs weight calculation using the back propagation method, a weight storage unit 90, and a weight change amount storage unit 92.

In this way, by providing the learning adjuster 80 for each connection of the rule section pre-wire neuro 100, it is possible to determine whether or not the learning scheduler 52 of the learning instruction unit 50 changes the weight value of the connection for the rule section pre-wire neuro 100. It is possible to set a learning flag that determines the weight, and it is possible to perform weight learning only for connections for which the learning flag is valid, that is, the flag is turned on.

Furthermore, the flag set for the learning coordinator 80 sets the antecedent membership that can execute fuzzy rules for a general hierarchical network that does not have a network structure that can implement fuzzy rules as shown in FIG. It is also used when specifying a hierarchical network structure consisting of a function realization section, a rule section, and a consequent membership function realization section.

When a hierarchical network conforming to fuzzy rules is configured, the weight values are adjusted as follows when actually learning is performed by the pack propagation method.

In FIG. 23, the weight change information reading unit 84 of the learning adjuster 80 monitors whether the weight change amount has been written by the weight calculation unit 88 to the weight change amount storage unit 92 of the weight change unit 76. There is. When the weight change information reading unit 84 detects that the weight has changed, it notifies the weight change amount adjustment unit 86 of the fact. The weight change amount adjustment unit 86 checks the learning flag in the flag storage unit 82, and does nothing if the learning flag is on. That is, writing to the weight change amount storage section 92 is made valid. On the other hand, if the learning flag is off, the weight change amount adjustment section 86 changes the weight change amount storage section 9 of the weight change section 76.
2 is set to zero to disable the weight change. The hardware configuration of this learning scheduler is disclosed in Japanese Patent Application No. 63-2278.
No. 25, “Learning method for network configuration data processing device”.

[Processing contents of learning scheduler] Next, referring to the learning processing flow in Fig. 21,
The learning process of the rule section pre-wired neuron according to the device configuration shown in the figure will be explained.

First, when the learning scheduler 52 of the learning instruction unit 50 is started, in step Sl (hereinafter, "step" is omitted), the phase counter i indicating the progress of the learning phase is set to i.
= 1, the process proceeds to S2, and the neuron group numbers Gb and Gc of the learning phase i=1 are read out with reference to the phase/group correspondence table 56. Next, the process proceeds to S3, and the connection group numbers Cab and Cbc corresponding to the neuron group numbers Gb and Gc are read out with reference to the group-connection correspondence table 58. Next, the process proceeds to S4, where the connection group numbers Cab and Cbc are output and the learning adjuster 8
Turn on the learning flag of 0. Specifically, the learning scheduler 52 of the learning instruction unit 50 sends the connection group number Cab,
Output and read Cbe.

The learning instruction reading section 72 sends the reading result to the learning flag setting section 74.
The learning flag setting unit 74 sets the flag of the learning adjuster 80 provided on the connections belonging to the connection group numbers Cab and Cbc that received the learning instruction, that is, the input connections of the units of the antecedent membership function realization unit 10. The learning flag is set by instructing the storage unit 82 to turn on the flag.

Next, the process proceeds to S5, where the learning scheduler 52 issues a learning command to start learning processing. This learning process uses the control human power X prepared in the learning signal storage unit 54 and the control human power
Read out the teacher signal d that is the desired control output for
The control human power X is given to the rule section prewire neuro 100, and the teacher signal d is given to the weight change section 76 and the learning convergence determination section 60.
give to Then, the control output Y output from the rule unit prewire neuro 100 that receives the control human power X is taken into the weight change unit 76 and the learning convergence determination unit 60, and the weight change unit 7
In step 6, the weight value of the antecedent membership function realization unit 10 designated as the learning target in phase 1 is changed according to the pack propagation method, and the learning convergence determination unit 60 changes the control output Y for the teacher signal d. It is determined that the error has become equal to or less than the specified value, and the learning process is terminated.

When the learning process of phase 1 is completed based on the learning execution command in step S5, the phase counter i is incremented by one in S6 to become phase 2, and after checking the learning phase in S7, the process returns to S2 again, and the learning phase i is
= enters 2.

In this phase 2 learning process, neuron group numbers Gd, G corresponding to learning phase i=2 are set in S2.
At the same time, the neuron group number Gd is read out in S3.
, connection group numbers Ccd, Cde corresponding to Ge
Read out the wiring group numbers Ccd, Cde in S4.
The learning flag of the learning adjuster 80 provided in the connection belonging to the rule section, that is, the connection in the rule section, is set. Subsequently, in S5, a learning execution command is issued, and learning of the weight values of the rule section 12 in Phase 2 is executed according to the pack propagation method. When phase 2 learning is completed, the phase counter i is counted at 86.
is incremented by one, and in S7 it is determined that the learning phase has ended, and the series of processing ends.

[Principle of learning by pack propagation method] Next, the principle of weight learning by the pack propagation method of the present invention will be explained together with the hierarchical network.

FIG. 24 shows a general hierarchical neural network, which is composed of a type of node called a unit and internal connections with weights.

In Figure 24, the h layer is the input device, and the i layer is the intermediate layer.
layer, and the j layer as the output layer. Of course, there is only one i-layer as the intermediate layer, but the same is true even if there are a plurality of intermediate layers. Each unit has the configuration shown in FIG. 4, and the calculations performed in the unit are as shown in equations (1) and (2) above for the basic unit 20. In the pack propagation method, the weight value Wih and threshold value θi of the hierarchical network are adaptively and automatically adjusted and learned using error feedback. As is clear from equations (1) and (2) above, it is necessary to adjust the weight value Wih and the threshold value θi at the same time, but this work is a difficult work that interferes with each other.

Therefore, in addition to the normal unit, the applicant of this application has
A unit whose output value is always 1 is provided in the h layer on the input side,
For the input connection from this unit with an output value of 1, the threshold value θj
By assigning as a weight value, the threshold θi becomes the weight value W
A pack propagation method has been proposed in which the threshold value θi can be treated as a weight value by incorporating it into i h (Japanese Patent Application No. 62-333484).

By doing this, the above equations (1,) and (2) become
n+=Σy ahWIh (II
)yp+=1/ (1+exp(-Xp+))
(12), and the threshold value θi does not appear.

By expanding from equations (11) and (12) above, we get the following (13)
) (14) is obtained.

x, sweat @ y, 1w, l
(13)y,,+1/(1+exp(-xpI))
(14) However, j: unit number W of the j layer, ,: weight value X+ of the internal connection between the i-j layers, +: sum of products y□ of inputs from each unit of the i layer to the 3rd unit of the j layer : Output from the 3rd unit of the j layer for the input signal of the p-th pattern The weight changing unit 76 in FIG.
When receiving the output cover turn Vp+ output from the output layer j and the teacher pattern apl, which is the signal to be taken by the output cover turn Vp+ when is presented, the difference value (dp+ Vp+) between the two is calculated, and then δp+=yp+ (1-yp+) (dp+-yp+)
(15) is calculated. Subsequently, Δw,, m = εΣδp + V D l + ζΔWz
(+-+) (16) However, ε: Learning constant ζ: Momentum t: Update amount ΔW of the weight value between the i layer and the j layer according to the number of learning times.

Calculate (1). Here, the reason for adding the weight value update amount determined in the previous update cycle in the second term on the right side of equation (16) is to speed up learning.

Subsequently, the weight value changing unit 76 uses Δδ21 calculated by the above equation (15) to calculate δ, , = 7p, (1y□)蓓δp+W++(f−1)
(17), then Δw+h(+) =εΣδpi)'ph+ζΔWlh(
+-1) (I8) The update amount Δw+h(+) of the weight values of the h layer and the i layer is calculated according to the following.

Subsequently, the weight change unit 76 determines the weight value for the next update cycle according to the update amounts calculated by the equations (16) and (18). The output pattern y□ for the input pattern for learning the update of the above weight values is the teacher pattern apl.
The learning process is repeated so that a weight value that matches is obtained.

On the other hand, when the units constituting the hierarchical network in FIG. 24 are linear function units, the equation (14) is y. ,=xp, (20)
It is expressed as The above (1
Equations 5) and (17) are δpI= d p+ V p+
(21) δ, l = copy δp, w,, (t −D,
It appears as (22).

[Weight Learning Process of the Present Invention] FIG. 25 is a process flow diagram of weight learning performed by the pack propagation method for the rule section pre-wire neuro of the present invention shown in FIG. 5A.

The parameters of each part in this learning processing flow are
It is defined as shown in Figure 6.

FIG. 26 schematically shows the hierarchical structure of the rule section pre-wire neuron in FIG. 5A. As shown, it has a six-layer structure from the first layer to the sixth layer.
The layer is made up of sigmoid function units, and the rest are made up of linear function units.

Here, the fuzzy inference value output by the realization of the fuzzy rule by the pre-wired neuron in the rule section, that is, the control output value, is y6, and the learning process in one phase is performed from the 6th layer to the 5th layer, the 4th layer, and the 3rd layer. Layer, second layer, first layer, and so on.

The progress of learning for each layer is indicated by an i counter.

That is, the i counter is set in the i layer 6 in the initial state of each phase.

To further simplify the explanation, Wl, l-1 and ΔW11 represent a matrix, the size of which is (number of neurons in the i-th layer) x (number of neurons in the i-1th layer). Furthermore, δi and yi represent vectors whose size is the same as the number of neurons in the i-th layer.

Based on the parameters of each part shown in FIG. 26, the weight learning process shown in FIG. 2 will be explained by dividing it into phase 1 and phase 2.

■Phase 1 In Phase 1, the antecedent membership function realization part 1
Only learn weights of 0. That is, the learning flags of the first and second layers belonging to the antecedent membership function realizing unit 10 in FIG. It has been reset to off.

First, a counter i for setting a learning target layer is initialized using Sl. Specifically, a counter i is initialized to i=6. With this initial setting, the sixth layer is designated as the learning target. Next, the process proceeds to 82, where it is determined whether the unit is a sigmoid function unit. Since the 6th layer is a linear function unit, the process proceeds to 814, where it is determined whether or not it is the final layer.Since the 6th layer is the final layer, the process proceeds to 315, where the teacher signal d obtained at that time and the control Using the output y, the difference value δ6 is determined according to a formula similar to the formula (21) above.

Next, the process proceeds to S6, where it is determined whether the unit is to learn weights. At this time, the 6th layer is excluded from the learning target, so
Specifically, since the learning flag of the learning adjuster 80 provided in the connection is off, the process proceeds to S8 and the weight update amount Δ
The process advances to S9 with W65=0, and in this case, the weight value is not updated. Next, the process proceeds to 812, where it is determined whether the counter i is the final learning layer, that is, the E-1st layer, and at this time, i=6.
Therefore, the process proceeds to S13, decrements the counter i by one, sets i=5, returns to S2, and starts learning the next fifth layer.

Since the fifth layer is also a linear function unit, 82
Proceeds to 84, but since it is not the final level, it goes to 816.
Then, the difference value δ5 is calculated according to a formula similar to the formula (22) above.
Calculate it.

Next, proceed to 86 to determine whether it is a learning target,
Since the fifth layer belonging to the consequent membership function realization unit 14 is excluded from the learning target, the process proceeds to S8 and sets the weight update amount ΔW 54 = 0, and therefore the weight update at 89 is not performed. The above processing for the fifth layer is repeated until the final unit, and the process proceeds to the fourth layer. Since the fourth layer is a linear function unit belonging to the rule section 12 and is excluded from the learning target, the difference value δ4 is calculated at 816 in the same way as the fifth layer.
Ask for only.

When the fourth layer processing is completed, the process proceeds to the third layer. Third
Since the layer is a sigmoid function unit, the process proceeds from S2 to 83, and since it is not the final stage, the process proceeds to 85, where the difference value δ3 unique to the sigmoid function unit is determined using the same calculation formula as the above equation (17). Since the third layer belonging to this rule section 12 is also excluded from the learning target, the process proceeds from S6 to 88, where the weight update amount ΔW3□=0, and the weight update in S9 is not performed.

When the third layer processing is completed, the process proceeds to the second layer.

The second layer belongs to the antecedent membership function realizing unit 10 that is the learning target in Phase 1, and is a linear function unit.

Therefore, the process proceeds from S2 to 814, and since it is not the final stage, the difference value δ2 is calculated in S16. In S6, it is determined that weight learning is to be performed, and the process proceeds to S7. The weight update amount ΔW21 is calculated according to the following. Next, in S9, the weight value W is calculated according to a formula similar to the formula (19) above.
Update 2+. Thereafter, the process is repeated in the same manner up to the final unit of the second layer, and then the process proceeds to the first layer.

The first layer belongs to the antecedent membership function realizing unit 10 and is therefore a learning target, and is composed of sigmoid function units. Therefore, the difference value δ1 is calculated in S5 via steps 82 and 83, it is determined that it is a learning target in S6, the weight update amount ΔWIO is calculated in S7, and the weight is updated in S9. This process is repeated until the final unit of the first layer, and when the final unit is determined in S10, the process proceeds to 312, and at this time, the counter i matches the learning end layer E=1, so the series of processes ends.

■Phase 2 In this phase 2, the weights of the rule section 12 are learned. For this reason, the learning flag of the learning adjuster 80 provided at the connection between the third layer and the fourth layer belonging to the rule section 12 in FIG. 5-6
The layer's training flag has been reset to off.

Weight learning in this phase 2 is executed with the value of the learning final layer E at 812 in FIG. 25 set to the value E of the third layer belonging to the rule section 12 = 3.

The sixth and fifth layers are the same as Phase 1. When the processing of the fourth layer belonging to the rule section 12 is entered, the weight update of 89 is effectively performed. The weights are updated in the same way for the third layer, and when the third layer is completed, the i counter matches i=3 and the learning end layer E=3, so the series of processing ends.

[Initial Setting of Weights Using Random Numbers] Next, another embodiment of the learning method of the present invention will be described with reference to the learning operation flow diagram of FIG. 27.

In the learning process shown in FIG. 2, the weights Cab, C of the antecedent membership function realizing unit 10 performed in step S1 are
Regarding the initial setting of bc, the present invention is characterized in that weight values are initially set randomly using random numbers. In this case, if there is some knowledge that is previously held, settings may be made based on that knowledge.

The initial setting of the weight values for the rule section 12 and the consequent membership function realization section 14 in steps S2 and S3 is performed based on the introduction of previously held knowledge, as in the embodiment of FIG.

In this way, in step S6, the function of the neural network is improved by randomly initializing the weight values of the antecedent membership function realizing unit 10, where the weights are finally learned in phase 1, using random numbers. You can expect efficient learning processing that brings out the best possible results.

[Other Notes: The above embodiment takes as an example the rule section pre-wire neuron that realizes fuzzy rules targeting two control human forces Xn and Xn, as shown in FIG. 5A, but the control Of course, the number of inputs may be 1 or 3 or more as appropriate.

In addition, in the above embodiment, y=1/ (1+e-wx'') shown in Figure 6.7.
(23) was taken as an example, but as shown in FIG. 28, the unit outputs of two sigmoid function units The function shape of (b) may also be realized. The unit output y in this case is y== 1/(1-e-WIX″11) 1/(+
-e-W2X″#2).

Furthermore, regarding the consequent membership function realization unit 14 of the present invention, after the weight values are initially set by introducing knowledge that is held in advance, learning by the pack propagation method is not performed. It does not necessarily need to be configured with a neuronetwork such as this, but may be a hardware configuration. This point is determined by the center of gravity determining element output device 40
The same applies to

[Effects of the Invention] As explained above, according to the present invention, the initial value of the weight value is After making the settings, by performing weight learning only for the antecedent membership function realization part, the antecedent membership function realization part can be efficiently utilized while effectively utilizing the weights of the rule part and the consequent membership function realization part. The weights can be adjusted. In addition, by learning the weights of the rule part after completing the weight learning of the antecedent part membership function realization part,
Efficient weight learning can be performed by effectively utilizing the weights of the learned antecedent membership function realization unit and the initialized consequent membership function realization unit.

Furthermore, by initializing the weights using random numbers in the antecedent membership function realization part or using a combination of pre-existing knowledge and random numbers, it is possible to maximize the learning function of the neutral network. can.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram of the principle of the present invention; Fig. 2 is a learning operation flow diagram of the present invention; Fig. 3 is a configuration diagram of an embodiment of the rule section pre-wired neuron of the present invention; Fig. 4 is an explanatory diagram of the basic unit; FIG. 5A is a specific example configuration diagram of the rule section prewire neuron of the present invention; FIG. 5B is an explanatory diagram of the neuron group and connection group of FIG. 5A; FIG. 6 is an antecedent section membership of the present invention. FIG. 7 is an explanatory diagram of the membership function of the antecedent part; FIG. 8 is an explanatory diagram of the initial setting of the rule section of the present invention; FIG. 9 is an explanatory diagram of the rule section unit of the present invention; Figure 10 is an explanatory diagram of fuzzy rule L1 in Figure 9; Figure 11 is an explanatory diagram of fuzzy rule L2 in Figure 9.
Explanatory diagram; FIG. 12 is an explanatory diagram of fuzzy rule L3 in FIG. 9; FIG. 13 is an explanatory diagram of fuzzy rule L4 in FIG. 9;
Figure 4 is an explanatory diagram of fanz rule L5 in Figure 9; Figure 15 is an explanatory diagram of fuzzy rule L6 in Figure 9; Figure 16 is an explanatory diagram of fuzzy rule Hiff in Figure 9; An explanatory diagram of fuzzy rule L8; FIG. 18 is an explanatory diagram of the initial settings of the consequent membership function implementation unit of the present invention; FIG. 19 is an explanatory diagram of the initial settings of the center of gravity calculation implementation unit of the invention; An explanatory diagram of the settings of the phase/group correspondence table and the group/connection correspondence table; Fig. 21 is a flowchart of the learning process of the present invention; Fig. 22 is a diagram of the device configuration of the learning process of the present invention; Fig. 23 is a diagram of the configuration of the learning process of the present invention; Fig. 24 is an explanatory diagram of a general hierarchical neural network; Fig. 25 is a weight learning process flow diagram of the present invention; Fig. 26 shows the parameters of each part in Fig. 25. Explanatory diagram; FIG. 27 is an operation flow diagram showing another embodiment of the present invention; FIG. 28 is an explanatory diagram of another membership function used in the present invention. In the figure, 10: Antecedent membership function implementation unit 12: Rule unit 14 Consequent membership function implementation unit 16, input unit 18: Center of gravity calculation implementation unit 20: Basic unit 22: Multiplication unit 24: Addition unit 26: Threshold processing unit 2B-128-232-1~32-4.38-1~38
-9.40-1.40-2: Linear function unit 30-1 to 3G-234-1 to 34-8.94-1.9
4-2: Sigmoid function unit 40: Center of gravity determining element output device 42: Center of gravity calculation device 50: Learning instruction unit 52: Learning scheduler 54: Learning signal storage unit 56: Phase/group correspondence table 58 Niger loop/phase correspondence table 60: Learning Convergence determination section 70: Learning unit 72: Learning instruction reading section 74: Learning flag setting section 76: Weight changing section 80: Learning adjuster 82: Flag storage section 84: Weight change information reading section 86: Weight change amount adjustment section 88: Weight calculation section 90: Weight storage section 92: Weight change amount storage section 96: Subtraction unit 100: Rule section Pre-wire neuro Figure 1 Output value of the name child sword period Ronu J Ding 2 A U steam map unit F11 1000 7 N 1 <Nup between Yoshimata Todoroki 1t BFI origin Fig. 7 Imauri Ming's rules static unit et al. Charming decoy Fig. 9 (a) Unknown Ru; Fuzzy rule Ll, normal XfSS) 0RX2ISS) THEN Y
(SS) L2:IF Xl(S S) ANDX2(S
S) THEN Y(LA)L3:TFX[LA) A
NDX2 (S s> THEN y<5s) L4:TF
X1 (LA) Omen ZSS) Forceps EN Y (LA) L5:
rFXl(SS) 0Rx2(LA) T HEN y
(ss)L6:rF XI(S s> AND !2
(LA) TT(EN y (LA)L7:IFXl(
LA) ANDXILA) TT(EN y(SS)L
8:W XI(LA) ORXza-A) TT(EN
Y (LA) (b) 1RU 訃 i 連 Ｇａ Ｇａｎ Ｇａｎ 明 fig. 58 c'J ('J ■ II II II -１１)－１１１C%JC'J αこゝ---Y－ --) Menkuゝ---■---' Prisoner 111)-1-J C%JC'%JOO 111)-1-J ('J('J ■ 1I II 1 11- i---J O〕 ?Toy ← Toi ← The weight of the attack TX Kartamami Shunka's Kito Ki Ha Vice Ema, Niwo Akito 4 Figure 19 Phase: Group correspondence table (a) (b) Title figure

Claims (6)

[Claims] (1) It has a hierarchical network consisting of an antecedent part membership function realization part (10), a rule part (12) consisting of one or more layers, and a consequent part membership function realization part (14), and the rule part (12) does not internally connect all units with the antecedent membership function realization unit (10) in the previous stage and/or the consequent membership function realization unit (14) in the subsequent stage, but only some parts according to the control rule. In a learning method of a rule section pre-wired neuro that is internally coupled and outputs one or more control operation quantities (Y) corresponding to input control state quantities (X_1, X_2,...X_n), The antecedent membership function realization unit (10), the rule unit (12), and the consequent membership function realization unit (14)
a first step of initially setting weight values for each of; the antecedent membership function realization unit (10) using learning data;
A second step of learning only the weights of the rule section (12); and a third step of learning only the weights of the rule section (12) using learning data after the second step is completed. How to learn pre-wired neuro. (2) Initial setting of the weight values in the first process is performed by using the weight values created based on knowledge held in advance by the antecedent membership function realization unit (10), the rule unit (12), and the 2. A learning method for a rule part prewired neuron according to claim 1, characterized in that initial settings are made in each of the subject membership function realization units (14). (3) The initial setting of the weight value in the first step is to set an initial weight value created using random numbers in the antecedent membership function realization unit (10), and to 2. A learning method for a rule section prewire neuron according to claim 1, wherein the subject membership function realizing section (14) is initially set with weight values created based on knowledge held in advance. (4) In setting the initial weight values of the antecedent membership function realization unit (10) in the first step, weights that can be set based on knowledge held in advance are initialized using this knowledge, and other weights are set using random numbers. 2. The method of learning a rule section pre-wired neuron according to claim 1, wherein the initial setting is performed using a pre-wired neuron. (5) At the start of learning in the second process, a learning flag provided for each connection of the antecedent membership function realization unit (10) is turned on, and the rule unit (12) and the consequent membership The rule according to claim 1, characterized in that a learning flag provided for each connection of the function realization unit (14) is turned off, and only the weight value of a connection having a learning flag in an on state is optimized by a learning process. How to learn part prewire neuro. (6) At the start of learning in the third process, the rule section (1
In addition to turning on the learning flag provided for each connection in 2), the learning flag provided for each connection of the antecedent membership function realization unit (10) and the consequent membership function realization unit (14) is turned on. 2. A learning method for a rule section pre-wired neuron according to claim 1, wherein only the weight values of connections having learning flags that are turned off and turned on are optimized through learning processing.