JPH0554013A - Learning device for neural network - Google Patents

Abstract

PURPOSE:To execute efficient learning of a neural network. CONSTITUTION:The learning device consists of a learning means X2 which allows the neural network to learn prescribed input/output relations, an error calculating means X3 which calculates learning error, an error deciding means X4 which deciding the extent of learning error, a frequency deciding means X5 which decides the frequency in learning in the case of learning error not smaller than a prescribed value, a learning error display means X6 which displays learning error, a continuation/stop command means X7 which commands continuation or stop of learning with a prescribed frequency at the time of completion of learning with the prescribed frequency, and a learning control means X8 which terminates learning in the case of learning error smaller than the prescribed value and continues learning with the designated prescribed frequency at the time of giving the continuation command and terminates learning at the time of giving the stop command. Since a worker recognizes the progress condition of learning and can obtain the learning result relatively to the learning time, he can command continuation or termination of learning.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a learning device for a neural network. More specifically, the present invention relates to a device that effectively completes learning by instructing whether to continue or end learning depending on the progress of 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 learns the coupling coefficient of the neural network in advance so that an optimum output can be obtained for each of a plurality of discrete inputs, and for any input It is a circuit network that allows a proper output to be directly obtained.

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

[0004]

This neural network is learned by using a large number of input data and optimum values of outputs 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 output data approaches the teacher data is executed for all sets of input data and teacher data. Is one learning operation.

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. .. The number of learnings until the coupling coefficient approaches a certain value depends on the input / output characteristics to be learned,
It reaches hundreds of thousands to millions of times. Therefore, if the termination of the learning operation is determined by the coupling coefficient approaching a certain value, it takes a long time for learning.

In an apparatus using a neural network to calculate a processing condition, normally, in the process of using the neural network, input data and teacher data which is an optimum output corresponding to the input data are accumulated. Incidentally, when the output of the neural network does not show an appropriate output, the neural network is appropriately trained based on the accumulated data.

[0006] In this case, every time learning is required, hundreds of thousands to several millions of learning operations are executed, and it takes a long time to complete the learning. Therefore, there is a problem that the numerical control device is occupied by the learning calculation during the learning and other processing cannot be performed. Further, there is a problem that the calculation of the processing conditions using the neural network cannot be started until the learning of the neural network is completed. As described above, the learning operation for a long time causes a decrease in workability.

The present invention has been made to solve the above problems, and an object thereof is to efficiently round up learning operations.

[0008]

As shown in FIG. 7, the present invention stores a large number of sets of input data and teacher data in a learning device for making a neural network learn based on input data and teacher data. Data storage means X
1, learning means X2 for sequentially correcting the coupling coefficient of the neural network so as to output corresponding teacher data for a large number of input data, and allowing the neural network to learn a predetermined input / output relationship; An error calculating means X3 for calculating a learning error representing the degree of progress, and an error determining means X for judging whether or not the learning error is smaller than a predetermined value.
4, the number of times determining means X5 for determining whether or not the number of times of learning reaches a predetermined number, the learning error display means X6 for displaying the learning error, and the number of times determining means for determining that the predetermined number of times of learning is completed. In this case, further, the continuation presence / absence command means X7 for instructing whether or not to continue the learning a predetermined number of times, and if the error determining means determines that the learning error is smaller than the predetermined value, the learning is terminated and continued. When the continuation instruction is given by the presence / absence instruction means, the learning means is activated to continue the learning for the designated number of times, and when the continuation presence / absence instruction means gives the discontinuation instruction, the learning is ended. The learning control means X8 is provided.

[0009]

The learning means executes the learning operation by correcting the coupling coefficient based on the input data and the teacher data stored in the data storage means. As the learning calculation progresses, a learning error representing the degree of learning progress is calculated by the error calculating means, and the error determining means determines whether or not the learning error is smaller than a predetermined value. If the learning error is smaller than the predetermined value, the learning control means ends the learning. If the learning error is larger than the predetermined value, the number-of-times determining means determines whether or not the number of times of learning reaches the predetermined value. The learning error is displayed by the learning error display means. The operator deciphers the displayed learning error and continues the learning or the learning error converges to zero to an acceptable level,
When the learning can be discontinued, the continuation presence / absence instruction means gives a discontinuation instruction, and when continuation learning is necessary, a continuation instruction is given. As a result, when the continuation command is given, the learning control means activates the learning means to further execute the next predetermined number of learning calculations.

[0010]

As described above, when the learning error is larger than the predetermined value and the learning operation is completed a predetermined number of times, the variation tendency of the displayed learning error is observed, and the instruction to continue or abort the learning is issued. Can be given. Therefore, when the convergence of the learning error is delayed, the learning operation is not unnecessarily repeated, and efficient learning is achieved.

[0011]

【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 has a control program area 2 in which a control program for managing the storage of input data and teacher data is stored.
1 and a neural network area 22 in which a calculation program of the neural network is stored and a learning program area 23 in which a program for learning the neural network is stored. RAM3
In the input data storage area 31 and the teacher data area 3 for storing the input data and the teacher data, respectively.
2. A coupling coefficient area 33 for storing the coupling coefficient of the neural network is formed. The CPU 1 is also connected to the CPU 1 via the input / output interface 5 with a keyboard 4 for giving an instruction as to whether learning should be continued or aborted, and a CRT 6 for displaying the relationship between the learning error and the learning frequency.

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 in 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] I ⁱ _j = ΣW ^i-1 _k, ⁱ _j · O ^i-1 _k + V ⁱ _j (2) ^k

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

Here, 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, 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.

[0016]

[Equation 4]

Next, at 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.

[0018]

[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.

[0019]

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

[0020]

[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 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.

[0022]

## _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.

[0024]

[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. Learning of Neural Network This neural network uses RO for initial learning.
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 3, respectively.
It is stored in 2.

In step 200 of FIG. 3, the data number i is set to the initial value 1, and the output element number j (teaching 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.

[0027]

[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 Zigmoid 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 determined whether or not learning data has been calculated for all output elements. If the determination result is NO, in step 208 the element number j is incremented by 1 and 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, at 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, the coupling coefficient is

[0031]

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.

[0033]

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

[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 pieces of 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.

When it is determined in step 216 that learning has been completed for all n input data and teacher data,
In step 220, the learning error ΔL is calculated. The learning error ΔL is calculated by the following equation.

ΔL = (W ₁ + W ₂ + ... + W _n ) / n (16)

W _i = (δ _i1 ² + δ _i2 ² + ... + δ _ig ² ) ^1/2 (17)

Δ _ij = O _ij −e _iJ (18) where W _i is the magnitude of the output error with respect to the input data D _i , and δ _ij is the j-th component of the output error with respect to the input data D _i , that is, the output layer. It is the output error of the j-th element of.
O _ij is the j-th component of the output for the input data D _i , and e _iJ is the j-th component of the teacher data for the output. n is the number of input data or teacher data, and g is the number of elements in the output layer. Therefore, the learning error ΔL represents the average value of the output errors regarding the n pieces of input data.

Next, at step 222, it is judged if the learning error ΔL is smaller than a predetermined value. If the learning error ΔL is smaller than the predetermined value, it means that the predetermined learning effect has been obtained, and thus the learning ends. If it is determined that the learning error ΔL is larger than the predetermined value, step 2
The process shifts to 24, and it is determined whether or not the number of times of learning 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 learning number has reached the predetermined number, the CRT 6 displays the relationship of the learning error ΔL with respect to the learning number as shown in FIG. The operator looks at the predicted portion (painted portion) of this display and judges whether the learning has considerably converged, whether the learning effect is still left, or the like. When the worker determines that the learning effect may not be improved even if the learning is continued for a long time and the learning may be terminated, the worker operates the keyboard 4 to give a discontinuation command. On the contrary, when the worker determines that the learning effect is improved by continuing the learning even if the time is further taken, the operator gives a continuation command.

In step 228, the CPU 1 reads the continuation command or the non-continuation command from the keyboard 4, and in the next step 230, it determines whether or not the command is a continuation command. If it is a continuation command, the predetermined learning number for determining the next learning number determined in step 224 is updated, the process returns to step 200, and the above-described learning calculation is repeated toward the next predetermined learning number.

In this way, the worker can see the convergence tendency of the learning effect on the way and further determine whether the learning should take time or not. Is possible. Therefore, since learning is not continuously executed unnecessarily for a long time, effective learning is achieved.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a calculation procedure of the neural network according to the embodiment.

FIG. 3 is a flowchart showing a learning procedure of the neural network according to the embodiment.

FIG. 4 is a program diagram showing a configuration of a learning device 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 an explanatory diagram showing the relationship between the learning error displayed by the CPU and the number of times of learning.

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

Claims (1)

[Claims] 1. A learning device for learning a neural network based on input data and teacher data, comprising: a data storage unit 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 that the corresponding teacher data is output, and learning means for learning the neural network to a predetermined input / output relationship, and an error for calculating a learning error representing the degree of progress of learning. A calculation unit, an error determination unit that determines whether or not the learning error is smaller than a predetermined value, a number determination unit that determines whether or not the number of times of learning reaches a predetermined number, and a learning error display that displays the learning error. And the number of times determination means determines that the predetermined number of learnings have been completed, whether to continue the predetermined number of learnings. If the continuation presence / absence command means for instructing whether or not the learning error is determined to be smaller than a predetermined value by the error determination means, the learning is terminated, and if the continuation instruction is given by the continuation presence / absence command means. A learning control means for activating the learning means, continuing the learning a predetermined number of times, and terminating the learning when a discontinuation instruction is given by the continuation presence / absence instruction means. Neural network learning device.