JPH0581230A - Learning device for neural network - Google Patents

Abstract

PURPOSE:To properly execute the learning of a neural network(NN). CONSTITUTION:In the case of adding new input data and teacher data and executing updated learning in order to change the I/O characteristics of the NN, the new input data are compared with all old input data already stored (300). When the same old data exist, two corresponding teacher data are displayed (304) and either one of them is selected (306). When the old teacher data are selected, the new input data and the new teacher data are not added. When the new teacher data are selected, teacher data corresponding to the same old input data are deleted (310) and the new input and teacher data are added (312). Since the same input data do not exist in data to be used for learning, I/O characteristics obtained by properly approximating the combination of the input data and the teacher data are learned.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a neural network update learning device. Specifically, when it is necessary to change the input / output characteristics of the neural network, it is evaluated whether a set of new input data and new teacher data should be added to the learning data for changing the input / output characteristics. By doing so, the present invention relates to a learning device in which the input / output characteristics are not inappropriately changed by update learning.

[0002]

2. Description of the Related Art A neural network is known as a circuit network that directly realizes a causal relationship that is difficult to theoretically analyze by a learning effect of a coupling coefficient. That is, the neural network modifies the coupling coefficient of the neural network in advance so that the optimum output can be obtained for each of a plurality of discrete inputs, and the neural network is appropriate for any input. It is a circuit network that can directly obtain various outputs.

Such neural networks are applied in many fields, and also in the field of machine tools.
It is used when computing the optimum machining conditions for the machining required from many setting conditions.

This neural network is trained by using a large number of input data and optimum output values corresponding to the input data, that is, teacher data. For one set of teacher data corresponding to certain input data, an operation of correcting all coupling coefficients in a direction in which the output data approaches the teacher data is performed on all sets of the input data and the teacher data. Thus, one learning calculation is performed.

By repeating this learning operation a number of times, all the coupling coefficients are sequentially corrected in the direction in which the corresponding teacher data is output for all the input data, and finally approach a certain value. .. As a result of such learning, the neural network comes to exhibit a predetermined input / output characteristic. However, in the process of using this neural network for many input data, the output result may not be appropriate.

In such a case, it is necessary to correct the input / output characteristics of the neural network, and for this reason, the set of new input data and new teacher data necessary for correcting the input / output characteristics is used for update learning. To be added to the data.

[0007]

However, in modifying the input / output characteristics of the neural network, a set of input data and teacher data desired to be learned by an operator or the like is given. In this case, generally, only the teacher data is corrected corresponding to the input data already given. In such a case, if the new input data and the new teacher data are added to the data storage means as they are, the two input data become completely the same even though the teacher data are different.

However, if the neural network is trained to learn the input / output characteristics by the backpropagation method with the data in which the two input data are equal to each other despite the different teacher data, the properness of the set of the input data and the teacher data is obtained. Learning is not executed, and proper learning is not executed for other sets of data.

That is, the input / output characteristics of the neural network learned so far are destroyed, and the input / output characteristics do not approximate the set of data given for the update learning.

The present invention has been made to solve the above problems, and an object thereof is to update and learn the input / output characteristics of a neural network by adding new input data and new teacher data. This is to make it possible to perform appropriate learning that closely approximates the set of input data and teacher data provided.

[0011]

According to the present invention, as shown in FIG. 7, in a learning device for causing a neural network X5 to update and learn predetermined input / output characteristics based on input data and teacher data, the input data and teacher Data storage means X1 storing a large number of sets of data, and for a large number of input data, the coupling coefficient of the neural network X5 is sequentially corrected so that the corresponding teacher data is output, and the neural network X5 has a predetermined value. Learning means X2 for learning the input / output characteristics, and when new input data and new teacher data are added to the data storage means X1 to change the input / output characteristics of the neural network, the new input data is stored in the data storage means. Judgment is made by comparing all the input data stored in X1 to determine whether or not there is input data equal to the new input data. And means X3, when the input data is equal to the new input data is determined not to exist,
If it is determined that there is new input data and first data updating means X4 for adding new teacher data corresponding to the new input data to the data storage means, and input data equal to the new input data, Display means X6 for displaying two teacher data corresponding to the input data, selecting means X7 for selecting one of the two displayed teacher data, and new teacher data when selected.
The new teacher data and the corresponding new input data are added to the data storage means X1, and the teacher data and the corresponding input data that have not been selected are deleted from the data storage means X1, and the new teacher data is selected. If not, the second data updating means X8 that does not add new teacher data and corresponding new input data to the data storage means is provided.

[0012]

When changing the input / output characteristics of the neural network, new input data and new teacher data are candidates for addition to the data storage means. First, the new input data is compared with all the input data stored in the data storage means to determine whether or not the same old input data as the new input data exists. If the same old input data does not exist, the new input data and the corresponding teacher data are added to the data storage means.

If the same old input data as the new input data exists in the data storage means, new teacher data corresponding to the new input data and teacher data corresponding to the old input data are displayed. The operator can select which teacher data to employ. When new teacher data is selected, teacher data corresponding to the same old input data is deleted from the data storage means, and the new input data and new teacher data are added to the data storage means. On the other hand, when old teacher data is selected, new input data and new teacher data are not added to the data storage means.

[0014]

As described above, according to the present invention, when new input data is added to the data storage means, if the same old input data exists, one of them is selected and the selected one is selected. Only the input data and the teacher data are adopted as the data for learning.

Therefore, in the data used for learning, it is excluded that the input data is the same even though the teacher data is different. Therefore, in learning the input / output characteristics of the neural network, the input data and the teacher data are excluded. Appropriate learning of input / output characteristics that closely approximates the data set is achieved.

[0016]

【Example】

1. Structure of learning device This device, as shown in FIG.
It is composed of a computer system composed of AM3. The ROM 2 stores a data update program area 21 in which a data update program for managing updates of input data and teacher data, a neural network area 22 in which a neural network operation program is stored, and a program for learning a neural network are stored. The learned program area 23 is formed.
Further, the RAM 3 is formed with an input data storage area 31 and a teacher data area 32 for respectively storing the input data and the teacher data to be accumulated, and a coupling coefficient area 33 for storing the coupling coefficient of the neural network. Also, CP
The U1 is connected via an input / output interface 5 to a keyboard 4 for giving a selection command of displayed teacher data and various commands and a CRT 6 for displaying two teacher data corresponding to the same input data.

2. Neural Network As shown in FIG. 1, the neural network 10 of this embodiment has a three-layer structure of an input layer LI, an output layer LO and an intermediate layer LM. The input layer LI has e input elements, the output layer LO has g output elements, and the intermediate layer LM has f output elements. A multilayered neural network is generally defined as a device that performs the following operations.

The output O ⁱ _j of the j-th element of the i-th layer is calculated by the following equation. However, i ≧ 2.

[Equation 1] O ⁱ _j = f (I ⁱ _j )
(1)

[Equation 2]

[Formula 3] f (x) = 1 / {1 + exp (-x)} (3)

Where V ⁱ _j is the bias of the j-th arithmetic element of the ^i- th layer, and W ^i-1 _k, ⁱ _j is the k-th element of the i-1 th layer and the j-th element of the i-th layer. The coupling coefficient between the th element, O ¹ _j , represents the output value of the j th element of the first layer. That is, since it is the first layer and the input is output as it is without performing any calculation, it is also the input value of the j-th element of the input layer (first layer).

Next, a specific calculation procedure of the neural network 10 having the three-layer structure shown in FIG. 1 will be described with reference to FIG. For the calculation of each element, R is referred to while referring to the coupling coefficient stored in the coupling coefficient storage area 33 of the RAM 3.
This is performed by executing a program stored in the neural network area 22 of the OM2. In step 100, the j-th element of the intermediate layer (second layer) is the output value O ¹ _j ( ^first element) from each element of the input layer (first layer).
(Input data of layer), input the equation (2) with the layer number and the first
The product-sum operation of the following equation, which is embodied using the number of elements of layers, is performed.

[0021]

[Equation 4]

Next, in step 102, the output of each element of the intermediate layer (second layer) is calculated by the following equation using the sigmoid function of the product-sum function value of the input values of equation (4).
The output value of the j-th element in the second layer is calculated by the following equation.

[0023]

[Equation 5] O ² _j = f (I ² _j ) = 1 / {1 + exp (-I ² _j )} (5) This output value O ² _j is input to each element of the output layer (third layer) It becomes a value. Next, in step 104, the output layer (third layer
The sum of products operation of the input values of each element of the (layer) is executed.

[0024]

[Equation 6] Next, in step 106, the output value O ³ _j of each element of the output layer is calculated by the sigmoid function, similarly to the equation (5).

[0025]

[Equation 7] O ³ _j = f (I ³ _j ) = 1 / {1 + exp (-I ³ _j )} (7)

3. Structures of input data and teacher data The data used for the update learning of the neural network is structured in a database as shown in FIG.
The input data are D _1, ..., D _n , and the corresponding teacher data are E _1, ..., E _n . The n pieces of input data and the teacher data are data accumulated in the process of actually using the initial learning of the neural network or the neural network after the initial learning. This input data is defined as follows. Consider the e pieces of data given to each of the e input elements as one set of data. And
An arbitrary m-th set of input data is represented by D _m , and the input data for the j-th input element belonging to the set is d _{m.
Expressed as mj} . D _m represents a vector, and d _mj is a component of the vector. That is, D _m is defined by the following equation.

[0027]

## _EQU8 ## D _m = (d _m1, d _m2, ..., d _me-1, d _me ) (8) Further, the n sets of input data are D _1, D _2, ..., D _n-1, D _n. It is represented by. Hereinafter, the input data group of all n sets is the input data group D
Is written. When the equation (4) is used for the input data D _m , the component d _mk is substituted into O ¹ _k of the equation (4).

Similarly, E _1, ... _, E _n are defined as follows. For the output layer Lo, consider the teacher data for the output from each of the g output elements as a set of data. Then, an arbitrary m-th set of teacher data is represented by E _m , and teacher data for the j-th output element belonging to that set is represented by _em j. E _m represents a vector, e
_mj is the component of that vector. That is, E _m is defined by the following equation.

[0029]

[Equation 9] E _m = (e _m1, e _m2, ..., E _mg-1, e _mg ) (9) Further, n sets of teacher data are E _1, E _2, ..., E _n-1, E _n. It is represented by. Hereinafter, the teacher data group E for all n sets is the teacher data group E.
Is written.

4. Update of Data The neural network 10 is learned by the procedure described below so that the initial prescribed input / output characteristics can be obtained.
Then, when the neural network on which the initial learning is performed is actually used, when the output is inappropriate, the input data and the appropriate teacher data corresponding to the input data are newly added as data for learning. Added to. Then, update learning of the neural network is performed in order to correct the input / output characteristics using these data. Such update learning is sequentially performed. At the time of this update learning, it is judged whether adding new input data and new teacher data to the data for update learning does not violate the learning tendency of the neural network until now. Addition is performed, and update learning of the neural network is performed.

Next, the procedure will be described with reference to FIG. In step 300, new input data D _n
And all old input data stored in the input data storage area 31 of the RAM 3 are compared. Next, in step 302, it is determined whether or not there is old input data that is the same as the new input data. If it is determined that the same input data exists, a warning that the same input data exists is displayed on the CRT 6 in step 304, and two teacher data corresponding to the same input data are displayed on the CRT 6. Is displayed.

Looking at the display of the CRT 6, the worker decides which teacher data is adopted for the update learning, and selects one of the teacher data. In step 306, the teacher data selected by the designation from the keyboard 4 is determined.

Next, in step 308, it is judged whether or not the selected teacher data is new teacher data E _n . If the judgment result is YES, in step 310, it is the same as the new input data D _n . The old input data and the corresponding old teacher data are deleted from the input data storage area 31 and the teacher data storage area 32 of the RAM 3, respectively.

Next, in step 312, new input data D _n and corresponding new teacher data E _n are added to the input data storage area 31 and the teacher data storage area 32, respectively.

On the other hand, when the selected data is not new teacher data E _n , that is, old teacher data,
RA of new input data D _n and new teacher data E _n
The process of the CPU 1 ends without adding to M3.

In step 302, the old input data identical to the new input data D _n is input data area 3
If it is determined that the data does not exist in step 1, step 312
And the new input data D _n and the corresponding new teacher data E _n are respectively stored in the input data storage area 31.
Is added to the teacher data storage area 32.

The data in the input data storage area 31 and the teacher data storage area 32 are updated by the above processing procedure. Therefore, during the update learning of the neural network, the same input data is not stored in the input data storage area 31, so that the appropriate learning of the neural network is achieved.

5. Learning of Neural Network Next, the processing procedure of update learning of the neural network will be described. This neural network is RO
Learning is performed by executing the program of the procedure shown in FIG. 3 stored in the learning program area 23 of M2.
The learning of the coupling coefficient is performed by the well-known backpropagation method. This learning is repeatedly executed with respect to a large number of input data regarding various events so that each output becomes each optimum teacher data. These input data and teacher data are stored in the input data storage area 31 and the teacher data storage area 32, respectively, by the procedure described above.

In step 200 of FIG. 3, the data number i is set to the initial value 1, and the output element number j (teacher data component number j) is set to the initial value 1. Next, in step 202, the i-th input data D _i and the i-th teacher data E _i are extracted from the input data storage area 31 and the teacher data storage area 32. next,
In step 204, the learning data Y of the element corresponding to that component of the output layer is calculated by the following equation.

[0040]

[Formula 10] Y ³ _j = (e _{ij -O} ³ _j ) f ^' (I ³ _j ) (10) However, in Y ³ _j , O ³ _j , and I ³ _j , the data number i is omitted. .. f ^' (x) is the derivative of the sigmoid function. Also, I ³
_j is obtained by substituting each component of the input data D _i into O ¹ _k of the equation (4) to obtain I ² _k with respect to all the elements in the intermediate layer, and substituting I ² _k into (5). obtains an output O ² _k for all elements, obtained by substituting the O ² _k in (6) with respect to all the k. Further, O ³ _j is obtained by substituting I ³ _j into the equation (7).

Next, in step 206, it is judged whether or not the learning data has been calculated for all the output elements. If the judgment result is NO, the element number j is incremented by 1 in step 208, Returning to 204, the learning data regarding the next output element is calculated.

When it is determined in step 206 that the calculation of the learning data for all output elements is completed, step 21
At 0, learning data Y 1 regarding an arbitrary r-th element in the intermediate layer is calculated by the following equation.

[Equation 11] The calculation of such learning data is executed for all the elements in the intermediate layer.

Next, in step 212, each coupling coefficient of the output layer is corrected. The correction amount is calculated by the following equation.

Δω ² _i, ³ _j (t) = P · Y ³ _j · f (I ² _i ) + Q · Δω ² _i, ³ _j (t-1) (12) where Δω ² _i, ³ _j (t) is the amount of change in the t-th calculation of the coupling coefficient between the j-th element in the output layer and the i-th element in the intermediate layer. Further, Δω ² _i, ³ _j (t-1) is the previous correction amount of the coupling coefficient. P and Q are proportional constants. Therefore,

[0044]

The corrected coupling coefficient is obtained by W ² _i, ³ _j + Δω ² _i, ³ _j (t) → W ² _i, ³ _j (13).

Next, in step 214, the coupling coefficient of each element of the intermediate layer is corrected. The correction amount of the coupling coefficient is obtained by the following equation, as in the case of the output layer.

[0046]

Δω ¹ _i, ² _j (t) = P · Y ² _j · f (I ¹ _i ) + Q · Δω ¹ _i, ² _j (t-1) (14) Therefore,

[Equation 15] The corrected coupling coefficient is obtained by W ¹ _i, ² _j + Δω ¹ _i, ² _j (t) → W ¹ _i, ² _j (15).

Next, at step 216, it is judged whether or not one learning is completed for the n input data and the teacher data to be learned. If learning for all input data has not been completed, the process proceeds to step 218 to read the next input data and the teacher data corresponding to the input data from the input data storage area 31 and the teacher data storage area 32. The data number i is incremented by 1, and the component number j is set to the initial value 1. Then, the process returns to step 202, and the above learning is executed using the next input data and the teacher data.

If it is determined in step 216 that learning has been completed for all n input data and teacher data,
It is determined whether or not the number of learning times has reached a predetermined value. When the number of times of learning has not reached the predetermined number (for example, 10,000 times), the process returns to step 202, and the learning calculation up to the next predetermined number of times is repeated. On the other hand, when it is determined that the number of times of learning has reached the predetermined number of times, the above learning calculation is completed.

In the above embodiment, the data storage means is composed of the input data storage area 31 and the teacher data storage area 32 of the RAM 3, and the learning means is the CPU 1 and R.
It is constituted by the learning program area 23 of the OM 2, and its function is achieved by the processing procedure of FIG. Further, the judging means and the first data updating means are constituted by the data updating program area 21 of the CPU 1 and the ROM 2, and the functions thereof are achieved by the steps 300, 302 and 312 of FIG. 6, respectively. Further, the display means is the CPU 1, RO
It is composed of the data update program area 21 of M2 and the CRT 6, and its function is achieved in step 304 of FIG. The selection means is composed of the CPU 1, the data update program area 21 of the ROM 2 and the keyboard 4, and the function thereof is achieved in step 306 of FIG.
The second data updating means is composed of the CPU 1 and the data updating program area 21 of the ROM 2, and its function is achieved by steps 308, 310 and 312 of FIG.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing a configuration of a neural network used in a learning device according to a specific embodiment of the present invention.

FIG. 2 is a flowchart showing a calculation procedure of a neural network.

FIG. 3 is a flowchart showing a learning procedure of a neural network.

FIG. 4 is a block diagram showing a configuration of a learning device according to a specific example of the present invention.

FIG. 5 is a configuration diagram showing a data configuration of a database having input data and teacher data used for learning of a neural network.

FIG. 6 is a flowchart showing a processing procedure for adding input data and teacher data by the CPU.

FIG. 7 is a block diagram showing the concept of the present invention.

[Explanation of symbols]

10 ... Neural network LI ... Input layer LM ... Intermediate layer LO ... Output layer 1 ... CPU (learning means, determination means, first data updating means, second data updating means, display means, selecting means) 2 ... ROM 3 ... RAM (Data storage means) 4 ... Keyboard (selection means) 9 6 ... CRT (display means) Steps 300, 302 ... Judgment means Step 304 ... Display means Step 306 ... Selection means Step 312 ... First data update means, Second data update Means Steps 308, 310 ... Second data updating means

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Shiho Hattori 1-1, Asahi-cho, Kariya city, Aichi prefecture Toyota Koki Co., Ltd.

Claims (1)

[Claims] 1. A learning device for causing a neural network to update and learn predetermined input / output characteristics based on input data and teacher data, comprising: a data storage means for storing a large number of sets of the input data and the teacher data; Learning means for sequentially correcting the coupling coefficient of the neural network so as to output corresponding teacher data for the input data of, and learning the predetermined input / output characteristics in the neural network, and the learning means. When new input data and new teacher data are added to the data storage means to change the input / output characteristics of the neural network, all the new input data stored in the data storage means are added. By comparing with the input data, it is determined whether or not the input data equal to the new input data exists. Determining means that determines the new input data and new teacher data corresponding to the new input data when the determination means determines that there is no input data equal to the new input data. First data updating means to be added to the storage means, and display means for displaying two teacher data corresponding to the input data when the determining means determines that there is input data equal to the new input data Selecting means for selecting one of the two teacher data displayed by the display means, and when the new teacher data is selected by the selecting means, the new teacher data and the corresponding new teacher data are selected. Input data is added to the data storage means, and the teacher data and the corresponding input data not selected are input to the data storage means. If the new teacher data is not selected and the new teacher data is not selected, new teacher data and corresponding new input data are not added to the data storage means.
A neural network learning device having data updating means.