JP6917004B2 - Evaluation device, evaluation method and its program - Google Patents

Description

The present invention relates to an evaluation device, an evaluation method, and a program thereof.

Conventionally, it has been known to control a system by using a neural network technique. For example, Patent Document 1 discloses an elevator system that uses a neural network to select an optimum car to be moved from a plurality of cars to a landing.

Further, Cited Document 2 discloses a method of performing additional learning on a neural network in order to reduce erroneous recognition in a pattern recognition device using a neural network.

Japanese Unexamined Patent Publication No. 2005-255289 Japanese Unexamined Patent Publication No. 9-62648

In a control system using neural network technology, additional learning of the neural network may be performed in order to improve control performance and the like. For example, at the time of the first learning, the learning data A randomly selected from the data obtained in the period from time 1 to time 2 in FIG. 11 is used, and then the data obtained in the period from time 2 to time 3 in FIG. 11 is used. It is conceivable that additional learning is performed using the learning data B randomly selected from.

In such a case, the learning data B may include inappropriate learning data due to the random selection of the learning data. When additional learning is performed using the additional learning data B including inappropriate learning data in this way, the control performance of the neural network after the additional learning deteriorates as compared with the neural network before the additional learning. There is.

Further, when a special event occurs in the period from time 2 to time 3, the neural network after the additional learning may be specialized for the special event. As a result, the control performance of the neural network after the additional learning may be deteriorated as compared with the neural network before the additional learning.

Therefore, an object of the present invention is to provide an evaluation device, an evaluation method, and a program thereof that can deal with a case where the control performance of the learning module deteriorates after additional learning.

The evaluation device for evaluating the second learning module obtained by additionally learning the first learning module according to one aspect of the present invention includes a learning goal receiving unit that receives a learning goal to be achieved by the second learning module, and at least learning. Whether or not the second learning module has achieved the learning goal by using the evaluation unit that evaluates the evaluation items included in the goal and generates the evaluation data and the learning goal and the evaluation data. It includes a determination unit for determining, and a learning module selection unit for acquiring a third learning module different from the second learning module, at least based on the learning objective, when it is determined that the learning target is not achieved.

According to this aspect, if the control performance of the learning module does not achieve the learning goal after the additional learning, it is possible to avoid continuing to use the learning module that did not achieve the learning goal. The reliability of the system used can be improved. For example, by avoiding the use of a learning module that does not achieve the learning goal, it is possible to prevent a decrease in the processing accuracy of the entire system. In addition, when the learning goal is not achieved, at least based on the learning goal, a control device that controls using the learning module and an evaluation are performed by a configuration in which a learning module different from the learning module that did not achieve the learning goal is acquired. Since it is not necessary to hold a plurality of learning modules in any of the evaluation devices, the hardware resource for recording the learning modules can be minimized.

In the evaluation device, the third learning module may be the first learning module. According to this aspect, since the control of the system can be continued by using the learning module before the performance deterioration, it is possible to prevent the stability of the system using the learning module from being lowered.

The evaluation device may further include a factor estimation unit that estimates the factors that the learning module could not achieve the learning target and generates factor improvement data related to the learning data used for the additional learning of the learning module. According to this aspect, useful information can be obtained from the learning module whose performance has deteriorated. By utilizing such information, it is possible to reduce the learning data and processing required for improving the accuracy of the learning module, and it is possible to reduce the load on the CPU.

In the evaluation device, the learning module selection unit may generate a learning instruction based on the factor improvement data. According to this aspect, it is possible to make the learning module perform efficient additional learning that can overcome the problems of the learning module whose performance has deteriorated. By using the learning module with this efficient additional learning in the system, it is possible to improve the performance deterioration due to the update of the learning module in the shortest time.

In the evaluation device, the factor estimation unit may output the estimated factor, receive the user input regarding the estimated factor, and generate the factor improvement data based on the user input. According to this aspect, it is possible to generate information on performance deterioration factors based on more reliable information.

The evaluation method for evaluating the second learning module obtained by additionally learning the first learning module according to another aspect of the present invention includes a step of accepting a learning goal to be achieved by the second learning module and at least a learning goal. The evaluation item included in is evaluated by the second learning module, and the process of generating the evaluation data and the learning target and the evaluation data are used to determine whether or not the second learning module has achieved the learning target. It includes a step and a step of acquiring a third learning module different from the second learning module, at least based on the learning goal, when it is determined that the learning goal is not achieved.

The program according to another aspect of the present invention is a program for causing a computer to evaluate a second learning module obtained by additionally learning a first learning module, and is a learning goal to be achieved by the second learning module. The second learning module evaluates at least the evaluation items included in the learning goal and generates the evaluation data, and the second learning module uses the learning goal and the evaluation data. When it is determined that the learning target is not achieved, the process of determining whether or not the above is achieved and the process of acquiring a third learning module different from the second learning module are executed.

According to the present invention, it is possible to provide an evaluation device, an evaluation method, and a program thereof that can deal with a case where the control performance of a learning module deteriorates after additional learning.

It is a figure which shows the network configuration of the learning system which concerns on embodiment of this invention. It is a figure which shows the physical structure of the learning apparatus which concerns on embodiment of this invention. It is a figure which shows the physical structure of the additional learning control apparatus which concerns on embodiment of this invention. It is a functional block diagram of the learning apparatus which concerns on embodiment of this invention. It is a functional block diagram of the additional learning control device which concerns on embodiment of this invention. It is a flowchart which shows the flow of the additional learning process executed by the additional learning control device which concerns on embodiment of this invention. It is a flowchart which shows the flow of the additional learning process executed by the additional learning control device which concerns on embodiment of this invention. This is an example of a screen output by the additional learning control device according to the embodiment of the present invention. It is a figure which shows the network configuration of the learning system which concerns on another embodiment of this invention. It is a figure which shows the physical structure of the evaluation apparatus which concerns on embodiment of this invention. It is a figure which shows the time series at the time of selecting a training data.

Embodiments of the present invention will be described with reference to the accompanying drawings. It should be noted that the following embodiments are for facilitating the understanding of the present invention, and are not for limiting and interpreting the present invention. Further, the present invention can be modified in various ways as long as it does not deviate from the gist thereof. Further, those skilled in the art can adopt an embodiment in which each element described below is replaced with an equal one, and such an embodiment is also included in the scope of the present invention.

(Network configuration)
The network configuration of the learning system 1 according to the predetermined embodiment of the present invention will be described with reference to FIG. The learning system 1 includes a learning device 10, an additional learning control device 20, one or a plurality of sensors 30, and a storage device 40. The learning device 10 is connected to the additional learning control device 20, one or a plurality of sensors 30 and the storage device 40 via the communication network N. The communication network N may be either a wired communication network or a wireless communication network configured by a wired or wireless line, and may be the Internet or a Local Area Network (LAN).

The learning device 10 learns the learning module based on the learning data stored in the storage device 40, and stores the learned module in the storage device 40. The learning device 10 according to the present embodiment includes a learning module, but the learning module may be provided in a device separate from the learning device 10.

The learning module includes a unit of dedicated or general-purpose hardware or software having learning ability, or a combination of one unit of the hardware or software. The learning module that performs the learning includes those that have already learned something from the learning data and those that have not been learned. Here, the learning ability means an ability that can improve the processing ability of a certain task based on the experience obtained from the learning data.

The additional learning control device 20 uses the learned module to output output data according to the characteristics of the input data. The additional learning control device 20 according to the present embodiment acquires a learned module or a duplicate of the learned module from the learning device 10 and sets it as a learning module. The additional learning control device 20 can evaluate the output data output using the set learning module and output the evaluation data. The additional learning control device 20 can determine whether or not the currently set learning module has achieved the learning goal by using the learning goal and the evaluation data to be achieved by the set learning module. When it is determined that the learning target has not been achieved, the additional learning control device 20 may acquire, for example, a previously set learning module from the learning device 10 and set it as a learning module. The additional learning control device 20 has a functional configuration of the evaluation device 60 described later, and substantially includes the evaluation device 60.

The duplicate of the learned module includes one unit of dedicated or general-purpose hardware or software capable of reproducing the function of the learned module, or a combination of one unit of the hardware or software.

A copy of the trained module does not necessarily have the ability to learn. Further, the configuration of the trained module and the configuration of the duplicate of the trained module do not necessarily have to match. A replica of the trained module also includes a learning module obtained by so-called distillation. That is, the duplicate of the trained module includes the other trained learning module obtained by training another learning module having a structure different from that of the trained module so as to maintain the function of the trained module.

Here, the other learning module may have a simpler structure than the learned module and may be more suitable for deployment, and for learning the other learning module, the output of the learned module may be obtained. Data may be used. In addition, the duplicate of the trained module can be obtained by changing the regularization method to prevent overfitting, changing the learning rate of backpropagation, and changing the weight coefficient update algorithm in the learning process of the learning module. Includes trained modules to be.

Further, acquiring a learned module or a duplicate of the learned module means acquiring information necessary for reproducing the function of the learned module in the additional learning control device 20. For example, when the training module includes a neural network, acquiring the trained module or a duplicate of the trained module means at least the number of layers of the neural network, the number of nodes for each layer, and the weight parameter of the link connecting the nodes. Acquiring information about the bias parameter for each node and the functional form of the activation function for each node.

The sensor 30 may be any of a physical quantity sensor for detecting a physical quantity, a chemical quantity sensor for detecting a chemical quantity, and an information sensor for detecting information, but the sensor 30 is not limited to these, and may include any sensor. Physical quantity sensors are, for example, cameras that detect light and output image data and video data, heart rate sensors that detect a person's heartbeat and output heart rate data, and blood pressure sensors that detect a person's blood pressure and output blood pressure data. It also includes vital sensors such as a body temperature sensor that detects a person's body temperature and outputs body temperature data, and also includes a sensor that detects an arbitrary physical quantity and outputs an electrical signal. The stoichiometry sensor includes, for example, a gas sensor, a humidity sensor, an ion sensor, and other sensors that detect an arbitrary stoichiometry and output an electric signal. The information sensor includes, for example, a sensor that detects a specific pattern from statistical data, and also includes a sensor that detects arbitrary information.

The storage device 40 stores the sensing data output by the sensor 30. Further, the storage device 40 stores the learned module output by the learning device 10. Although the storage device 40 is shown as a single storage unit in FIG. 1, the storage device 40 may be composed of one or a plurality of file servers.

Although the learning device 10, the additional learning control device 20, and the storage device 40 are configured as separate bodies in FIG. 1, they may be configured as one. That is, the learning device 10, the additional learning control device 20, and the storage device 40 may all be integrally configured, and two of the learning device 10, the additional learning control device 20, and the storage device 40 may be selectively integrated. It may be configured. At this time, the elements of the learning device 10, the additional learning control device 20, and the storage device 40, which are integrally configured, are connected via an internal bus.

(Physical configuration: learning device)
The physical configuration of the learning device 10 according to the predetermined embodiment of the present invention will be described with reference to FIG. The learning device 10 includes a control unit 10a, a storage unit 10b, a communication unit 10c, an input unit 10d, and a display unit 10e. Each of these configurations is connected to each other via a bus so that data can be transmitted and received.

The control unit 10a includes a Central Processing Unit (CPU) corresponding to a hardware processor and a Random Access Memory (RAM) corresponding to a memory. The CPU expands the program stored in the storage unit 10b into the RAM, interprets and executes the program expanded in the RAM, and the control unit 10a functions as each unit of FIG. 4 to be described later.

The type of hardware processor is not limited to the CPU. For example, as a hardware processor, CPU, Graphics Processing Unit (GPU), Field-programmable Gate Array (FPGA), Digital Signal Processor (DSP), Application Specic Can be done. The RAM is a storage unit capable of rewriting data, and is composed of, for example, a semiconductor storage element. The RAM temporarily stores programs and data such as applications executed by the CPU.

The storage unit 10b is a non-volatile storage medium such as a Hard Disk Drive (HDD) or a Solid State Drive (SDD). The storage unit 10b stores programs and data executed by the CPU.

The communication unit 10c is a hardware interface that connects the learning device 10 to the communication network N.

The input unit 10d receives input from the user, and is composed of, for example, a keyboard, a mouse, and a touch panel.

The display unit 10e visually displays the processing result by the CPU, and is composed of, for example, a Liquid Crystal Display (LCD).

The learning device 10 may be configured by executing the learning program according to the present embodiment by, for example, the CPU of a general personal computer. The learning program may be stored and provided in a storage medium readable by a computer such as a RAM or a storage unit 10b, or may be provided via a communication network N connected by the communication unit 10c. These physical configurations are exemplary and do not necessarily have to be independent configurations.

(Physical configuration: additional learning control device)
The physical configuration of the additional learning control device 20 according to the predetermined embodiment of the present invention will be described with reference to FIG. Like the learning device 10, the additional learning control device 20 also has a control unit 20a including a CPU and a RAM, a storage unit 20b for storing data and the like, a communication unit 20c for connecting to the network N, and an input for receiving input from a user. It has a unit 20d, a display unit 20e, and the like. Each of these configurations is connected to each other via a bus so that data can be transmitted and received. The CPU expands the program stored in the storage unit 20b into the RAM, interprets and executes the program expanded in the RAM, and the control unit 20a functions as each unit of FIG. 5 described later.

The additional learning control device 20 may be configured by executing an additional learning control program by, for example, a CPU of a general personal computer. The additional learning control program may be stored and provided in a storage medium readable by a computer such as a RAM or a storage unit 20b, or may be provided via a communication network N connected by the communication unit 20c.

(Functional configuration: learning device)
With reference to FIG. 4, the functional configuration of the learning device 10 according to the predetermined embodiment of the present invention will be described. The learning device 10 includes a learning instruction receiving unit 101, a learning data acquisition unit 102, a learning control unit 103, a learning module 104, a learned module output unit 105, and a learned module extraction unit 106.

The learning instruction receiving unit 101 receives a learning instruction from the user via the input unit 10d or a learning instruction from the additional learning control device 20 via the communication unit 10c, and inputs the information included in the learning instruction to the learning data described later. It is handed over to the acquisition unit 102. In the present embodiment, the learning instruction includes learning data acquisition conditions, designation of input parameters, and the like. The learning data acquisition condition refers to a condition required to satisfy a learning instruction from a user among data that can be used as learning data for training the learning module 104. For example, among the data acquired by the sensor 30, the acquisition date and time may be specified. The input parameter refers to a factor that affects the control performance of the learned module among the information included in the learning instruction.

The learning data acquisition unit 102 receives the learning data acquisition condition and acquires the learning data from the storage device 40 based on the received learning data acquisition condition.

The learning control unit 103 trains the learning module 104 using the learning data acquired by the learning data acquisition unit 102. The learning control unit 103 completes learning based on the learning instruction received by the learning instruction receiving unit 101. The criterion for determining that the learning is completed may be, for example, the case where learning is performed using a predetermined number of learning data. Alternatively, learning may be completed when the control performance of the learned module satisfies the learning objective described later. When the learning is completed, the learning control unit 103 stores the learned module in the storage device 40. At this time, in the present embodiment, the learning control unit 103 stores the learned module in association with the learning module identifier and the learning data acquisition condition that can uniquely identify the learned module.

Here, it is desirable that the learned module stored in the storage device 40 is version-controlled so that its update history can be known. Version control may be performed by the learning module identifier itself or by the version information provided separately.

The learning module 104 is a module for realizing machine learning. Here, an example in which a neural network is applied as an example of the learning module 104 will be described. However, the neural network is only an example of the learning module 104, and the learning device 10 may apply another configuration as the learning module 104.

The learned module output unit 105 outputs the learned module and the learning module identifier to the outside such as the additional learning control device 20.

The learned module extraction unit 106 receives the learning module extraction condition, and acquires the learned module from the storage device 40 based on the received learning module extraction condition. In the present embodiment, the learning module extraction condition includes the currently set learning module identifier of the learning module and the extraction point. The extraction point includes information that can specify the learning module to be extracted, such as the date before the performance deterioration and the specification of the version that goes back from the currently set learning module. For example, the extraction point can be "before December 31, 2017" or "the version immediately before the currently set learning module".

(Functional configuration: Additional learning control device)
With reference to FIG. 5, the functional configuration of the additional learning control device 20 according to the predetermined embodiment of the present invention will be described. The additional learning control device 20 includes a learned module reception unit 201, a learning module 202, a learning target reception unit 203, an evaluation unit 204, a judgment unit 205, a learning module selection unit 206, a factor estimation unit 207, a control unit 208, and a database (DB). ) 209.

The learned module reception unit 201 receives the learned module and the learning module identifier, and sets the received learned module as the learning module 202. In the present embodiment, the learned module receiving unit 201 receives the learned module and the learning module identifier from the learned module output unit 105 of the learning device 10, and sets the received learned module as the learning module 202. The learned module reception unit 201 may receive the learned module from the storage device 40 and set it as the learning module 202. Here, an example in which a neural network is applied as an example of the learning module 202 will be described. However, the neural network is only an example of the learning module 202, and the additional learning control device 20 may apply another configuration as the learning module 202.

The learning goal receiving unit 203 receives the learning goal to be achieved by the learning module 202 via the input unit 20d, and stores the received learning goal in the DB 209. In this embodiment, the learning goal includes evaluation items and conditions. The evaluation item is an item for evaluating the learning module 202, and is, for example, an item used for determining the accuracy of the output data output by the learning module.

The conditions are, for example, for the evaluation item "number of external operations / day for the device controlled by the control unit 208", "the evaluation item is the reference value x or less", and "the evaluation item is immediately before". For the evaluation item "power consumption of the device controlled by the control unit 208 / month", "the evaluation item is less than or equal to the reference value y", and "the evaluation item is set immediately before". For the evaluation items "the number of times the value calculated by the learning module 202 exceeds the permissible change rate / month", "the evaluation item is less than or equal to the reference value z", and " The evaluation item is smaller than the learning module set immediately before. "

The evaluation unit 204 evaluates the learning module 202 using the sensing data stored in the storage device 40, generates evaluation data, associates it with the learning module identifier received by the learned module reception unit 201, and generates the evaluation. The data is stored in the DB 209. In the present embodiment, the evaluation unit 204 can evaluate at least the evaluation items included in the learning goal and generate evaluation data. The evaluation unit 204 may automatically perform the evaluation after a lapse of a certain period of time after the learning module 202 is set, or may separately receive an evaluation instruction from the user to perform the evaluation. For example, the evaluation unit 204 may perform the evaluation in response to accepting the learning goal. In addition to the evaluation items included in the learning goal, the evaluation unit 204 may evaluate preset evaluation items.

The determination unit 205 determines whether or not the currently set learning module has achieved the learning target by using the learning target received by the learning target reception unit 203 and the evaluation data stored in the DB 209. When it is determined that the learning goal has been achieved, the determination unit 205 ends the process.

When it is determined that the learning target has not been achieved, the learning module selection unit 206 transmits a learning module extraction condition to the learned module extraction unit 106 of the learning device 10 as described later, and the learning obtained as a response thereof. The completed module is set as the learning module 202. The learned module obtained by the learning module selection unit 206 transmitting the learning module extraction condition to the learned module extraction unit 106 is evaluated to achieve the learning goal again before being set as the learning module 202. May be good.

Further, the learning module selection unit 206 may instruct the factor estimation unit 207, which will be described later, to estimate the factors for which the learning goal could not be achieved, and obtain factor improvement data. After that, the learning module selection unit 206 may instruct the learning device 10 to perform re-additional learning using the factor improvement data obtained from the factor estimation unit 207.

The factor estimation unit 207 estimates the factors that could not achieve the learning goal when it is determined that the currently set learning module has not achieved the learning goal. In the present embodiment, the factor estimation unit 207 responds to the instruction from the learning module selection unit 206 and uses the learning target and the sensing data of the storage device 40 to determine the factor that could not achieve the learning target. presume. The sensing data of the storage device 40 is statistically processed in order to estimate the factors that failed to achieve the learning goal.

The control unit 208 controls using the value calculated by the learning module 202. In the present embodiment, a control unit that controls the system is shown as an example of controlling the system using the learning module. However, as will be described later, the embodiment of the present invention applies to various systems that execute processing using the learning module. Can be applied.

The learning target DB2091 and the evaluation data DB2092 are stored in the DB 209. In the present embodiment, the learning goal DB 2091 stores evaluation items and conditions included in the learning goal received by the learning goal reception unit 203. Further, in the evaluation data DB 2092, the evaluation data generated by the evaluation unit 204 and the learning module identifier of the learning module to be evaluated are stored in association with each other.

(Additional learning process)
[First Embodiment]
A first embodiment of the additional learning process executed by the additional learning control device 20 will be described with reference to the flowchart of FIG. The embodiment of the present invention can be applied to various systems that execute processing using a learning module, and the field thereof is not particularly limited, but in the following description, an air conditioning control system will be described as an example.

In the first embodiment, the control unit 208 is, for example, an air conditioning control unit. The learning module 202 calculates the room temperature set value by using the values such as the current room temperature, the outside air temperature, and the humidity output by the sensor 30, the known date and time, the volume of the room, and the like as input parameters. The control unit 208 of the additional learning control device 20 performs air conditioning control using the room temperature set value calculated by the learning module 202. Here, an embodiment in which the control unit 208 actually performs air conditioning control using the learning module 202 and the evaluation unit 204 evaluates the control result by the control unit 208 will be described. However, as an alternative, the additional learning control device 20 may include a simulation unit (not shown). In the alternative embodiment, the simulation unit performs air conditioning control using the learning module 202, and the evaluation unit 204 can evaluate the simulation result by the simulation unit.

Further, although an embodiment in which the additional learning control device 20 includes the learning module 202 and the control unit 208 will be described here, as an alternative, a control device separate from the additional learning control device 20 provides the learning module and the control unit. You may prepare.

As a result of learning performed in advance by the learning control unit 103 of the learning device 10, the storage device 40 is a neural network learned using the learning data for the period from January 1, 2010 to December 31, 2012. 0, the neural network 1 learned using the training data from January 1, 2010 to December 31, 2014, and the learning from January 1, 2010 to December 31, 2016. It is assumed that the neural network 2 learned using the data is stored. The neural network 0 is stored in the storage device 40 in association with the learning module identifier 0. The neural network 1 is stored in the storage device 40 in association with the learning module identifier 1. The neural network 2 is stored in the storage device 40 in association with the learning module identifier 2.

It is assumed that the additional learning control device 20 sets the neural network 1 as the learning module 202 until January 31, 2017. The storage device 40 stores the sensing data output by the sensor 30 at any time. That is, the storage device 40 contains sensing data regarding input parameters such as the current room temperature and outside temperature during the period during which the air conditioning control is performed using the neural network 1, the number of external operations for the room temperature set value, the amount of power consumption, and the like. Sensing data related to the evaluation items of is stored. Further, as a result of the evaluation previously performed by the evaluation unit 204, the evaluation data 0 of the neural network 0 and the evaluation data 1 of the neural network 1 are stored in the evaluation data DB 2092 of the additional learning control device 20.

In S601, the learned module reception unit 201 of the additional learning control device 20 receives the neural network and the learning module identifier, and sets the received neural network as the learning module 202. In the present embodiment, on February 1, 2017, the trained module reception unit 201 receives the neural network 2 and the learning module identifier 2 from the trained module output unit 105 of the learning device 10, and receives the neural network 2 Is set as the learning module 202. Since February 1, 2017, the control unit 208 has been performing air conditioning control using the room temperature set value calculated by the set neural network 2. During that time, as described above, the storage device 40 stores the sensing data output by the sensor 30 at any time during the period during which the air conditioning control is performed using the neural network 2.

In S602, the learning target receiving unit 203 of the additional learning control device 20 receives the learning target to be achieved by the learning module 202 via the input unit 20d, and stores the received learning target in the learning target DB2091 of the DB 209. In the present embodiment, on April 1, 2017, a learning goal is received from the administrator of the additional learning control device 20 via the input unit 20d, and the learning goal receiving unit 203 sends the received learning goal to the learning goal DB2091. Remember. The learning goal shall include the evaluation item "number of external operations / day for the room temperature set value" and the condition "less than the learning module set immediately before".

In S603, the evaluation unit 204 of the additional learning control device 20 evaluates the learning module 202, generates evaluation data, associates it with the learning module identifier received by the learned module reception unit 201, and associates the generated evaluation data with the DB 209. It is stored in the evaluation data DB 2092 of. In the present embodiment, in response to receiving the learning target in S602, the evaluation unit 204 uses the sensing data stored in the storage device 40 during the period when the air conditioning control is performed using the neural network 2. The neural network 2 is evaluated.

Here, the evaluation unit 204 may evaluate not only the evaluation items included in the learning goal received in S602 but also other preset evaluation items. With such a configuration, it is possible to cope with the case where the comparison condition with the previously set learning module is specified for different evaluation items in the future learning goal. In the present embodiment, the evaluation unit 204 evaluates not only the evaluation item "number of external operations / day for the room temperature set value" included in the learning target but also other preset evaluation items to generate evaluation data 2. Then, the evaluation data 2 is stored in the evaluation data DB 2092 in association with the learning module identifier 2.

Next, in S604, the determination unit 205 of the additional learning control device 20 uses the learning target received by the learning target reception unit 203 in S602 and the evaluation data stored in the evaluation data DB 2092, and is currently set for learning. Determine if the module has achieved its learning goals. When it is determined that the learning target has been achieved (S604: Yes), the additional learning control device 20 ends the process.

In the present embodiment, the determination unit 205 includes a learning goal including the evaluation item "number of external operations / day for the room temperature set value" and the condition "less than the learning module set immediately before", the evaluation data 1, and the evaluation. Using the data 2, it is determined whether or not the neural network 2 has achieved the learning goal. As a result, it is determined that the learning goal has not been achieved (S604: No), and the process proceeds to S605.

In S605, the learning module selection unit 206 of the additional learning control device 20 transmits the learning module extraction condition to the learned module extraction unit 106 of the learning device 10, and sets the learning module obtained as the response as the learning module 202. , End the process. In the present embodiment, the learning module selection unit 206 has the learning module extraction condition "learning module identifier of the currently set learning module: learning module identifier 2, extraction point: currently set learning" in the learned module extraction unit 106. The "previous version of the module" is transmitted, and the neural network 1 obtained as a response is set as the learning module 202.

[Second Embodiment]
In the first embodiment, an example of returning the neural network 2 determined not to achieve the learning goal to the version before the update, the neural network 1, has been described. In the second embodiment, an example in which the neural network is further trained when it is determined that the learning goal has not been achieved will be described.

FIG. 7 is a flowchart showing the flow of the additional learning process executed by the additional learning control device 20. In the second embodiment, the description of the matters common to the first embodiment will be omitted, and only the differences will be described. Since S701 to S705 in FIG. 7 are the same processes as S601 to S605 in FIG. 6, detailed description of these processes will be omitted.

In S706, the learning module selection unit 206 instructs the factor estimation unit 207 of the additional learning control device 20 to estimate the factors for which the learning target could not be achieved, and the factor estimation unit 207 achieves the learning target. Estimate the factors that could not be achieved. In the present embodiment, the factor estimation unit 207 estimates the factor for the evaluation item “number of external operations / day with respect to the room temperature set value” included in the learning target by using the sensing data of the storage device 40. For example, the factor estimation unit 207 estimates the factor using the sensing data in the period from January 1, 2010 to December 31, 2016, which is the period of the training data used in the neural network 2.

In the present embodiment, the factor estimation unit 207 uses a known algorithm, and for example, the evaluation item “number of external operations / day with respect to the room temperature set value” included in the learning goal is often greatly deviated from the values of other days. Find possible patterns for the day. For example, it is assumed that a pattern is found in which "the number of external operations / day with respect to the room temperature set value" is large in a specific period "from July 1, 2016 to July 31, 2016". It should be noted that the pattern is not limited to the period, but when the values of the outside air temperature, humidity, etc. included in the input parameters continue to be different from the predetermined values, the pattern that the "input parameter value" is large or small is used. You may find it. Further, when the volume of the room included in the input parameter is changed, the pattern that the "volume of the room" is changed may be found.

In S707, the factor estimation unit 207 determines whether or not an outlier pattern has been found. If no outlier pattern is found (S707: No), the additional learning control device 20 ends the process. On the other hand, when an outlier pattern is found (S707: Yes), the process proceeds to S708, and the factor estimation unit 207 may output the found outlier pattern. In the present embodiment, as shown in FIG. 8, the text 801 indicating the pattern of the found outliers “From July 1, 2016 to July 31, 2016, the number of external operations / day with respect to the room temperature set value is large. A factor confirmation screen including a performance deterioration factor radio button 802, a detailed setting button 803, and a decision button 804 for setting whether or not the pattern is a performance deterioration factor is output. The detailed setting button will be described later.

In S709, the factor estimation unit 207 determines whether or not the user input has been received. When the user input is received, the process proceeds to S710, and the factor estimation unit 207 generates factor improvement data based on the user input, and returns the generated factor improvement data to the learning module selection unit 206. It is not desirable to include the factor improvement data in the training data because it is selected as a performance deterioration factor by the performance deterioration factor radio button 802 among the outlier patterns output by the factor estimation unit 207 in S708. The indicated pattern is included.

In the present embodiment, the administrator of the additional learning control device 20 confirms that the indoor construction was carried out during the period from July 1, 2016 to July 31, 2016, and the screen output in S708. In the above, it is assumed that the performance deterioration factor radio button 802 is selected as the performance deterioration factor and the enter button 804 is pressed. As a result, the factor estimation unit 207 generates factor improvement data "excluding the period from July 1, 2016 to July 31, 2016" based on the received user input, and learns the generated factor improvement data. Return to the module selection unit 206.

In S711, the learning module selection unit 206 generates a learning instruction based on the factor improvement data, transmits the generated learning instruction to the learning instruction receiving unit 101 of the learning device 10, and ends the process. In the present embodiment, the learning module selection unit 206 sets the “learning data acquisition condition: from January 1, 2010 to December 31, 2016, and from July 1, 2016 to July 31, 2016. A learning instruction including "exclude" is transmitted to the learning instruction receiving unit 101.

In the present embodiment, when it is determined that the outlier pattern is found in S707, the factor improvement data is generated based on the user's input in S710, but the user's input is accepted by omitting S708 to S710. Instead, the factor estimation unit 207 may generate factor improvement data based on the found outlier pattern.

By doing so, the learning device 10 can further train the neural network based on the received learning instruction. In the present embodiment, the learning device 10 can further train the neural network 1 by using the learning data excluding the period from July 1, 2016 to July 31, 2016.

[Third Embodiment]
In the second embodiment, when it is determined that the learning goal has not been achieved, data that is not desirable to be included in the learning data is excluded, and an example in which the neural network is further trained has been described. In the third embodiment, when it is determined that the learning target has not been achieved, data to be further trained is added to the neural network, and an example in which the neural network is additionally trained will be described.

A third embodiment will be described with reference to FIG. 7. Also in the third embodiment, the description of the matters common to those of the first embodiment and the second embodiment will be omitted, and only the differences will be described. Since S701 to S705 of FIG. 7 are the same processes as those in the first embodiment, detailed description of these processes will be omitted.

In S706, the factor estimation unit 207 estimates the factors for which the learning goal could not be achieved. In the present embodiment, it is assumed that a pattern is found in which the "number of external operations with respect to the room temperature set value" is large in a specific time zone "every Saturday from 8:00 to 9:00". It should be noted that the pattern is not limited to the period, but when the values of the outside air temperature, humidity, etc. included in the input parameters continue to be different from the predetermined values, the pattern that the "input parameter value" is large or small is used. You may find it. Further, when the volume of the room included in the input parameter is changed, the pattern that the "volume of the room" is changed may be found.

In S707, the factor estimation unit 207 determines whether or not an outlier pattern has been found. In the present embodiment, since the pattern of the outliers has been found, the process proceeds to S708, and the factor estimation unit 207 may output the pattern of the found outliers. In the present embodiment, the text 801 indicating the pattern of the found outliers "There are many external operations on the room temperature set value from 8:00 to 9:00 every Saturday", the performance deterioration factor radio button 802, the detailed setting button 803, and the detailed setting button 803. The factor confirmation screen including the decision button 804 is output.

In the present embodiment, the administrator of the additional learning control device 20 desires to set the room temperature setting value higher than that on weekdays because a smaller number of users use the room every Saturday than on weekdays. After confirming, on the screen output in S709, it is assumed that the performance deterioration factor radio button 802 is selected as the non-performance deterioration factor and the detailed setting button 803 is pressed. For example, by pressing the detailed setting button 803, the detailed setting screen is displayed, and the administrator may adjust the input parameters on this detailed setting screen. Alternatively, the additional learning control device 20 may adjust the input parameters.

In S709, the factor estimation unit 207 determines whether or not the user input has been received. In the present embodiment, it is assumed that the administrator adjusts the input parameters on the detailed setting screen and then presses the enter button 804. As a result, the process proceeds to S710, and the factor estimation unit 207 generates factor improvement data based on the user input, and returns the generated factor improvement data to the learning module selection unit 206. In the present embodiment, the factor improvement data includes the settings related to the input parameters specified on the detailed setting screen.

In S711, the learning module selection unit 206 generates a learning instruction based on the factor improvement data, transmits the generated learning instruction to the learning instruction receiving unit 101 of the learning device 10, and ends the process. In the present embodiment, the learning module selection unit 206 transmits a learning instruction including settings related to the input parameters specified on the detailed setting screen to the learning instruction receiving unit 101.

In the present embodiment, when it is determined that the outlier pattern is found in S707, the factor improvement data is generated based on the user's input in S710, but the user's input is accepted by omitting S708 to S710. Instead, the factor estimation unit 207 may generate factor improvement data based on the found outlier pattern.

By doing so, the learning device 10 can further train the neural network based on the received learning instruction. In the present embodiment, the learning device 10 uses the additional learning data when the pattern found by the additional learning control device 20 is not a performance deterioration factor but a pattern to be further trained by the neural network, and the neural network 2 Can be further learned.

[Another Embodiment]
As described above, in another embodiment, as shown in FIG. 9, the learning system 1 may include a control device 50 including a learned module reception unit 201, a learning module 202, and a control unit 208. good. In the present embodiment in which the learning system 1 includes the control device 50, the learning system 1 is an evaluation device including a learning target reception unit 203, an evaluation unit 204, a determination unit 205, a learning module selection unit 206, a factor estimation unit 207, and a DB 209. 60 is provided. As described above, the additional learning control device 20 includes the evaluation device 60.

As shown in FIG. 10, the evaluation device 60 receives input from a control unit 60a including a CPU and RAM, a storage unit 60b for storing data of DB209, a communication unit 60c for connecting to a network N, and an input from a user. It includes an input unit 60d, a display unit 60e, and the like. Each of these configurations is connected to each other via a bus so that data can be transmitted and received. The CPU expands the program stored in the storage unit into the RAM, interprets and executes the program expanded in the RAM, and the control unit 60a functions as each unit in FIG.

The evaluation device 60 may be configured by executing an additional learning program by, for example, a CPU of a general personal computer. The additional learning program may be stored and provided in a storage medium readable by a computer such as a RAM or a storage unit 60b, or may be provided via a communication network N connected by the communication unit.

The program that performs each process described in the present specification may be stored in a recording medium. By using this recording medium, the computer can be made to function as the evaluation device 60 or the additional learning control device 20 by installing the above program on the computer. Here, the recording medium in which the above program is stored may be a non-transient recording medium. The non-transient recording medium is not particularly limited, but may be, for example, a recording medium such as a CD-ROM.

Some or all of the above embodiments may also be described, but not limited to:

(Appendix 1)
An evaluation device for evaluating a second learning module in which a first learning module is additionally trained, which includes at least one memory and at least one hardware processor connected to the memory.
The hardware processor
Accepting the learning goals to be achieved by the second learning module,
At least the evaluation items included in the learning goal are evaluated by the second learning module, and evaluation data is generated.
Using the learning goal and the evaluation data, it is determined whether or not the second learning module has achieved the learning goal.
If it is determined that the learning goal is not achieved, a third learning module different from the second learning module is acquired, at least based on the learning goal.
Evaluation device.

(Appendix 2)
This is an evaluation method for evaluating the second learning module in which the first learning module is additionally learned.
At least one or more hardware processors accept the learning goals to be achieved by the second learning module.
The hardware processor evaluates the second learning module for at least the evaluation items included in the learning goal, generates evaluation data, and generates evaluation data.
Using the learning goal and the evaluation data, the hardware processor determines whether or not the second learning module has achieved the learning goal.
If it is determined that the learning goal is not achieved, a third learning module different from the second learning module is acquired, at least based on the learning goal.
Evaluation method.

1 ... Learning system, 10 ... Learning device, 10a ... Control unit, 10b ... Storage unit, 10c ... Communication unit, 10d ... Input unit, 10e ... Display unit, 101 ... Learning instruction reception unit, 102 ... Learning data acquisition unit, 103 ... learning control unit, 104 ... learning module, 105 ... learned module output unit, 106 ... learned module extraction unit, 20 ... additional learning control device, 20a ... control unit, 20b ... storage unit, 20c ... communication unit, 20d ... Input unit, 20e ... Display unit, 201 ... Learned module reception unit, 202 ... Learning module, 203 ... Learning target reception unit, 204 ... Evaluation unit, 205 ... Judgment unit, 206 ... Learning module selection unit, 207 ... Factor estimation unit , 208 ... Control unit, 209 ... DB, 2091 ... Learning target DB, 2092 ... Evaluation data DB, 30 ... Sensor, 40 ... Storage device, 50 ... Control device, 60 ... Evaluation device, 60a ... Control unit, 60b ... Storage unit , 60c ... communication unit, 60d ... input unit, 60e ... display unit, 801 ... text, 802 ... performance deterioration factor radio button, 803 ... detailed setting button, 804 ... enter button

Claims (7)

It is an evaluation device that evaluates the second learning module in which the first learning module is additionally learned.
A learning goal reception unit that receives learning goals to be achieved by the second learning module,
An evaluation unit that evaluates the second learning module and generates evaluation data for at least the evaluation items included in the learning goal.
Using the learning goal and the evaluation data, a determination unit for determining whether or not the second learning module has achieved the learning goal, and
Wherein when it is determined not to achieve learning objectives, evaluated and a learning module selection unit for acquiring third learning module different from the second learning module device. The evaluation device according to claim 1, wherein the third learning module is the first learning module. The claim further includes a factor estimation unit that estimates a factor that the second learning module could not achieve the learning goal and generates factor improvement data related to the learning data used for additional learning of the learning module. The evaluation device according to 1. The evaluation device according to claim 3, wherein the learning module selection unit generates a learning instruction based on the factor improvement data. The evaluation device according to claim 3, wherein the factor estimation unit outputs the estimated factor, receives a user input regarding the estimated factor, and generates the factor improvement data based on the user input. An evaluation method for evaluating a second learning module in which the first learning module is additionally learned, and a computer provided with a control unit is used.
The process of accepting the learning goals to be achieved by the second learning module, and
A process of evaluating the second learning module for at least the evaluation items included in the learning goal and generating evaluation data, and
A step of determining whether or not the second learning module has achieved the learning goal by using the learning goal and the evaluation data, and
Wherein when it is determined not to achieve learning objectives, evaluation method comprising a step of acquiring a third learning module different from the second learning modules. A program for causing a computer to evaluate a second learning module in which the first learning module is additionally learned.
The process of accepting the learning goal to be achieved by the second learning module,
At least the evaluation items included in the learning goal are evaluated by the second learning module, and evaluation data is generated.
A process of determining whether or not the second learning module has achieved the learning goal by using the learning goal and the evaluation data.
A program that executes a process of acquiring a third learning module different from the second learning module when it is determined that the learning goal is not achieved.

Images