TRAINING SUPPORT DEVICE, TRAINING SUPPORT METHOD, TRAINING SUPPORT PROGRAM, LEARNING DEVICE, LEARNING METHOD, AND LEARNING PROGRAM

Abstract

In order to generate a training menu while taking into account a state relating to training performed by a subject, a training support device (2) comprises a data acquisition unit (21) for acquiring state data indicating a state relating to the training, and a generating unit (22) for performing an optimization calculation using an objective function generated by inverse reinforcement learning using training data that indicate a training menu corresponding to the state relating to training, and thereby generating a training menu that corresponds to the state indicated by the state data.

Description

[Technical field]

[0001] The present invention relates to a training support device that supports training. [Background technology]

[0002] Technology to support training has been developed for some time. For example, the following Patent Document 1 discloses a technology for outputting a training menu based on a reference table or a mathematical model derived in advance from a statistical database based on customer data, etc. Specifically, Patent Document 1 describes calculating calories burned from the type and number of times of training performed by a subject, and returning a training menu that will allow the calories burned to reach a target value. Patent Document 1 also describes returning a training menu that takes into account psychological tendencies based on the results of a questionnaire. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2017-010486 Summary of the Invention [Problem to be solved by the invention]

[0004] The technology of Patent Document 1 has room for improvement in that it cannot take into account various conditions related to training when deciding a training menu. For example, it is possible that some of the training items included in the training menu generated by the technology of Patent Document 1 cannot be performed at the training facility used by the subject, or that the subject has difficulty performing the training due to physical limitations.

[0005] One aspect of the present invention has been made in consideration of the above problems, and one example of an objective thereof is to provide a technique for generating a training menu taking into account a state related to training. [Means for solving the problem]

[0006] A training support device according to one aspect of the present invention comprises a data acquisition means for acquiring status data indicating a status related to training performed by a subject, and a generation means for generating a training menu corresponding to the status indicated by the status data by performing an optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status related to the training.

[0007] A training support method according to one aspect of the present invention includes at least one processor acquiring status data indicating a status related to training performed by a subject, and generating a training menu corresponding to the status indicated by the status data by performing an optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status related to the training.

[0008] A training support program according to one aspect of the present invention causes a computer to function as a data acquisition means for acquiring status data indicating a status related to training performed by a subject, and as a generation means for generating a training menu corresponding to a status indicated by the status data by performing optimization calculations using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in that status related to the training.

[0009] A learning device according to one aspect of the present invention includes a data acquisition means for acquiring learning data indicating a training menu to be applied in a training state according to the state, and a learning means for generating an objective function for generating a training menu according to the state by performing inverse reinforcement learning using the learning data.

[0010] A learning method according to one aspect of the present invention includes at least one processor acquiring learning data indicating a training menu to be applied in a training state according to the state, and performing inverse reinforcement learning using the learning data to generate an objective function for generating a training menu according to the state.

[0011] A learning program according to one aspect of the present invention causes a computer to function as a data acquisition means for acquiring learning data indicating a training menu to be applied in a training state according to the state, and as a learning means for generating an objective function for generating a training menu in accordance with the state by performing inverse reinforcement learning using the learning data. Effect of the Invention

[0012] According to one aspect of the present invention, it is possible to generate a training menu taking into account conditions related to training. [Brief description of the drawings]

[0013] [Figure 1] 1 is a block diagram showing a configuration of a training support system according to a first exemplary embodiment of the present invention. [Diagram 2] FIG. 2 is a flow chart showing the flow of a learning method according to the first exemplary embodiment of the present invention. [Diagram 3] 1 is a flowchart showing a flow of a training support method according to an exemplary embodiment 1 of the present invention. [Figure 4] FIG. 11 is a diagram showing an overview of a training support method according to an exemplary embodiment 2 of the present invention. [Diagram 5] FIG. 11 is a block diagram showing an example of a configuration of a main part of a training support device according to an exemplary embodiment 2 of the present invention. [Figure 6] FIG. 11 is a diagram illustrating an overview of learning of an objective function according to an exemplary embodiment 2 of the present invention. [Figure 7] FIG. 13 is a diagram showing an example of generating a training menu including background music. [Figure 8]FIG. 13 is a diagram showing an example of a display screen for a training menu and background music. [Figure 9] 11 is a flowchart showing a flow of processing executed by a training support device according to an exemplary embodiment 2 of the present invention. FIG. [Figure 10] FIG. 1 is a diagram showing an example of a computer that executes instructions of a program, which is software that realizes the functions of each device according to each exemplary embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Example 1 A first exemplary embodiment of the present invention will be described in detail with reference to the drawings. This exemplary embodiment is a basic form of the exemplary embodiments described later.

[0015] (Training Support System) A training support system 3 according to this exemplary embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the training support system 3. The training support system 3 is a system for supporting training of a subject, and includes a learning device 1 and a training support device 2 as shown in the figure. The learning device 1 includes a data acquisition unit 11 and a learning unit 12. The training support device 2 includes a data acquisition unit 21 and a generation unit 22.

[0016] The data acquisition unit 11 acquires learning data that indicates a training menu to be applied in a given state according to the state related to training.

[0017] The learning unit 12 performs inverse reinforcement learning using the learning data to generate an objective function for generating a training menu according to the state.

[0018] The data acquiring unit 21 acquires condition data indicating the condition of the subject regarding the training performed by the subject.

[0019] The generating unit 22 generates a training menu according to the state indicated by the state data by performing optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the state related to the training. Note that the objective function used by the generating unit 22 may be generated by the learning unit 12 of the learning device 1 or may be generated by another device.

[0020] As described above, the learning device 1 according to this exemplary embodiment includes a data acquisition unit 11 that acquires learning data indicating a training menu to be applied in a state related to training, and a learning unit 12 that generates an objective function for generating a training menu according to the state by performing inverse reinforcement learning using the learning data. Therefore, the learning device 1 according to this exemplary embodiment can generate an objective function for generating a training menu according to the state, and therefore has the effect of enabling a training menu to be generated in consideration of the state related to training.

[0021] As described above, the training support device 2 according to this exemplary embodiment includes a data acquisition unit 21 that acquires state data indicating a state related to the training performed by the subject, and a generation unit 22 that generates a training menu according to the state indicated by the state data by performing optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the state related to the training. Therefore, the training support device 2 according to this exemplary embodiment has an effect of being able to generate a training menu in consideration of the state related to the training.

[0022] (Study Program) The above-mentioned functions of the learning device 1 can also be realized by a program. The learning program according to this exemplary embodiment causes a computer to function as a data acquisition means for acquiring learning data indicating a training menu to be applied in a training state, and as a learning means for generating an objective function for generating a training menu in accordance with the state by performing inverse reinforcement learning using the learning data. According to this learning program, since it is possible to generate an objective function for generating a training menu in accordance with the state, it is possible to obtain an effect that it is possible to generate a training menu taking into account the state related to training.

[0023] (Training Support Program) The above-mentioned functions of the training support device 2 can also be realized by a program. The training support program according to this exemplary embodiment causes a computer to function as a data acquisition means for acquiring status data indicating a status related to training performed by a subject, and as a generation means for generating a training menu corresponding to a status indicated by the status data by performing optimization calculations using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status related to the training. This training support program provides the effect of being able to generate a training menu in consideration of the status related to the training.

[0024] (Learning method flow) The flow of the learning method according to this exemplary embodiment will be described with reference to Fig. 2. Fig. 2 is a flow diagram showing the flow of the learning method. Note that the execution subject of each step in this learning method may be a processor included in the learning device 1, or a processor included in another device, or each step may be a processor provided in a different device.

[0025] In S11, at least one processor obtains learning data that indicates a training menu to be applied in a training-related state.

[0026] In S12, at least one processor performs inverse reinforcement learning using the learning data to generate an objective function for generating a training menu according to the state.

[0027] As described above, the learning method according to the present exemplary embodiment employs a configuration including: at least one processor acquiring learning data indicating a training menu to be applied in a training state according to the state; and performing inverse reinforcement learning using the learning data to generate an objective function for generating a training menu according to the state. Therefore, according to the learning method according to the present exemplary embodiment, since an objective function for generating a training menu according to the state can be generated, it is possible to obtain an effect that a training menu can be generated taking into account the training state.

[0028] (Flow of training support method) The flow of the training support method according to this exemplary embodiment will be described with reference to Fig. 3. Fig. 3 is a flow diagram showing the flow of the training support method. Note that the execution subject of each step in this training support method may be a processor included in the training support device 2, or a processor included in another device, or each step may be a processor provided in a different device.

[0029] In S21, at least one processor acquires status data indicative of a status related to the training performed by the subject.

[0030] In S22, at least one processor performs optimization calculations using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in a training state, thereby generating a training menu corresponding to the state indicated by the state data.

[0031] As described above, the training support method according to the present exemplary embodiment employs a configuration including: at least one processor acquires state data indicating a state related to the training performed by the subject; and performs optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the state related to the training, thereby generating a training menu corresponding to the state indicated by the state data. Therefore, the training support method according to the present exemplary embodiment has an effect of being able to generate a training menu taking into account the state related to the training.

[0032] Exemplary embodiment 2 (overview) The training support method according to this exemplary embodiment will be outlined with reference to Fig. 4. Fig. 4 is a diagram showing the outline of the training support method according to this exemplary embodiment (hereinafter referred to as this method).

[0033] As shown in the figure, the input data in this method includes training items selectable by the subject who will be performing the training, item properties indicating the characteristics of the items, user properties indicating the characteristics of the subject, and constraint conditions for generating a training menu. Of these, the item properties and user properties are status data indicating the status of the training performed by the subject. In this method, a training menu is generated that satisfies the constraint conditions and corresponds to the status indicated by the status data.

[0034] Specifically, the event property shown in FIG. 4 indicates the training events that the subject can select and the effects that can be expected from the training of each event. For example, the event property shown in FIG. 4 indicates that a training event called "Event 1" can be selected, and indicates that the muscle hypertrophy effect of this event is 80 and the effect of improving muscle output is 70. The method of evaluating the effect of training is not particularly limited, and any training effect evaluated by any evaluation method can be included in the event list. For example, the effect of improving calories burned or muscle endurance may be associated with each event. In addition to the effects, the exercise intensity, the part of the body to which the load is applied, etc. can also be included in the event property. The calories burned may be calories burned per unit time, or when the training time for the event is determined, it may be calories burned over the entire time.

[0035] 4 also shows the height and weight of the subject. The user property may be anything that indicates the characteristics of the subject, and may be, for example, the maximum training time per day, the minimum calories burned per day, the maximum exercise intensity per day, age, sex, occupation, exercise experience, and training goal. Note that the maximum training time per day may be set to a different time for each day of the week.

[0036] The constraint conditions shown in Fig. 4 indicate that the calorie consumption is equal to or greater than a target value, and that the total time required for training is within a set value. The constraint conditions can be set arbitrarily, and for example, a constraint condition can be set such as applying an even load to each part of the body in one week. Such constraint conditions can be set and changed freely by the subject.

[0037] In this method, a training menu is generated by performing optimization calculations using the input data described above and an objective function generated by inverse reinforcement learning using learning data that indicates a training menu to be applied in a given state according to the state of training.

[0038] The above objective function includes weights indicating the degree of importance to be attached to each aspect of evaluating the training menu. In the objective function shown in FIG. 4, the first aspect is "muscular hypertrophy", the second aspect is "muscular output", and the third aspect is "muscular endurance", and the weights of these aspects are α, β, and γ, respectively. The aspects can be automatically determined during learning, or the subject can be allowed to set and change the aspect. The aspect to be set may be anything related to the training menu, and for example, the compatibility of a combination of training events can be set as the aspect.

[0039] The output data, i.e., the training menu, shown in Fig. 4 indicates the training events to be performed on each day of the week and the order in which they are performed. For example, the training menu shown in Fig. 4 indicates that on Monday, training events 3, 2, and 5 are to be performed in that order. Note that this method is not limited to the example shown in Fig. 4, and any type of training menu can be generated. For example, it is possible to generate a training menu that specifies a combination of events without specifying the order in which the events are performed, or a training menu for one month at a time.

[0040] (Configuration of training support device) A training support device 2A according to this exemplary embodiment will be described with reference to Fig. 5. Fig. 5 is a block diagram showing an example of a main configuration of the training support device 2A. The training support device 2A is a device that supports the training of a subject by generating a training menu for the subject. The training support device 2A also has the function of the learning device 1 of the exemplary embodiment 1, that is, a function of generating an objective function including a weight value indicating the degree of importance to be attached to each aspect of evaluating the training menu.

[0041] As shown in the figure, the training support device 2A includes a control unit 20A that controls each unit of the training support device 2A, and a storage unit 21A that stores various data used by the training support device 2A. The training support device 2A also includes an input unit 22A that accepts input of various data to the training support device 2A, and an output unit 23A that allows the training support device 2A to output various data. Note that, although an example in which the output unit 23A is a display device that displays and outputs various data will be described below, the output unit 23A may output data in other output formats such as audio output or print output.

[0042] The control unit 20A includes a data acquisition unit 201, a generation unit 202, a search unit 203, and a learning unit 204. The storage unit 21A stores state data 211, an objective function 212, a training menu 213, and learning data 214. The search unit 203 will be described later in the section "Display of search keywords."

[0043] The data acquisition unit 201 acquires status data 211 indicating a status related to the training performed by the subject. The status data 211 may be anything that indicates a status related to the training performed by the subject. For example, the status data 211 may include information related to the training itself, such as the event property shown in FIG. 4, or information related to the subject himself, such as user property.

[0044] The data acquisition unit 201 may also acquire constraint conditions for generating a training menu. The data acquisition unit 201 also acquires learning data 214 indicating a training menu to be applied in a state related to training, according to the state. The method of acquiring these data is not particularly limited, and for example, the data acquisition unit 201 may acquire state data 211, constraint conditions, and learning data 214 input via the input unit 22A.

[0045] The generating unit 202 generates a training menu according to the state indicated by the state data acquired by the data acquiring unit 201. More specifically, the generating unit 202 performs optimization calculation using an objective function 212 generated by inverse reinforcement learning using learning data 214 indicating a training menu to be applied in a state related to training, thereby generating a training menu 213 according to the state indicated by the state data acquired by the data acquiring unit 201. A method of generating the training menu 213 will be described later in the section "Optimization Calculation".

[0046] The learning unit 204 uses the learning data 214 to generate an objective function 212 for generating a training menu according to a state. As described with reference to Fig. 4, the objective function 212 generated by the learning unit 204 indicates each viewpoint for evaluating the training menu, and includes a weight value indicating how much importance is attached to each viewpoint. A method for generating the objective function 212 will be described later in the section "Learning Objective Functions".

[0047] As described above, the training support device 2A according to this exemplary embodiment includes a data acquisition unit 201 that acquires state data 211 indicating a state related to training performed by a subject, and a generation unit 202 that performs optimization calculation using an objective function 212 generated by inverse reinforcement learning using learning data 214 indicating a training menu to be applied in a state related to the training, thereby generating a training menu 213 corresponding to a state indicated by the state data acquired by the data acquisition unit 201. Therefore, the training support device 2A according to this exemplary embodiment has an effect of being able to generate a training menu in consideration of a state related to training.

[0048] As described above, in the training support device 2A according to this exemplary embodiment, the data acquisition unit 201 acquires the constraint conditions for creating a training menu for a subject, and the generation unit 202 generates the training menu 213 that satisfies the acquired constraint conditions. Therefore, according to the training support device 2A according to this exemplary embodiment, in addition to the effects of the training support device 2 according to the exemplary embodiment 1, the effect of being able to generate the training menu 213 that satisfies a desired constraint condition can be obtained.

[0049] Furthermore, the training support device 2A according to this exemplary embodiment also functions as a learning device. That is, the training support device 2A includes a data acquisition unit 201 that acquires learning data 214 indicating a training menu to be applied in a state related to training, and a learning unit 204 that generates an objective function 212 for generating a training menu according to the state by performing inverse reinforcement learning using the learning data 214. According to the training support device 2A according to this exemplary embodiment, it is possible to generate an objective function 212 including a weight value indicating how much importance is attached to each aspect of evaluating the training menu, and therefore it is possible to generate a training menu 213 in consideration of the aspect that is emphasized by the subject of training.

[0050] (Learning the objective function) The learning of the objective function 212 by the learning unit 204 will be described with reference to Fig. 6. Fig. 6 is a diagram showing an overview of the learning of the objective function 212. The learning data 214 shown in Fig. 6 includes user properties and event properties as state data, as well as constraint conditions and a training menu.

[0051] The learning data 214 may be data indicating a training menu to be applied in a state according to the state related to the training. For example, the learning data 214 may be data indicating a training menu performed by a person who actually performed training and achieved remarkable results (hereinafter referred to as an expert). In this case, the user property in the learning data 214 is the user property of the expert, and the type property is the property of the training type performed by the expert. In addition, the constraint conditions of the learning data 214 may be constraint conditions when the expert created the training menu, and the training menu may indicate the training menu performed by the expert.

[0052] In the following, in this embodiment, the learning data 214 will be described in association with an expert. However, the learning data 214 is not limited to an expert. The learning data 214 may be a combination of data indicating a state related to training and data indicating a training menu to be applied in that state.

[0053] Furthermore, the learning data 214 does not necessarily have to be generated based on a training menu that has actually been performed. For example, typical condition data may be created, and an appropriate training menu corresponding to the typical condition data may be created to serve as the learning data 214. For example, if there is an appropriate training menu for a man in his twenties, the training event properties of the training events included in the training menu may be associated with user properties typical of a man in his twenties to serve as the learning data 214. Furthermore, constraint conditions may be included in the learning data 214 as necessary.

[0054] The learning unit 204 generates an objective function 212 including weight values ​​indicating the degree of importance to be attached to each viewpoint for evaluating a training menu by learning using a plurality of pieces of learning data 214 as described above, each having a different state. This learning can be said to be learning the intention of the expert in adopting a training menu included in a certain piece of learning data 214 when the expert is in a state indicated by state data included in the learning data 214. Note that the viewpoints can be set arbitrarily, as described above.

[0055] In learning, first, the learning unit 204 sets each weight value of the objective function 212 to an initial value. Next, the generating unit 202 generates a training menu according to the state indicated by the state data included in the learning data 214 by an optimization calculation using the objective function 212 with the weight values ​​set to the initial values. Then, the learning unit 204 updates the weight values ​​so that the difference between the training menu indicated in the learning data 214 and the training menu generated by the generating unit 202 becomes small. These processes are repeated until the difference between the training menu indicated in the learning data 214 and the training menu generated by the generating unit 202 becomes sufficiently small, whereby the learning of the objective function 212 is completed.

[0056] As a specific learning method, various methods used in general inverse reinforcement learning can be applied. For example, a maximum entropy inverse reinforcement learning method can be applied. In this case, the learning unit 204 expresses the probability distribution of the objective function using the maximum entropy principle, and learns the objective function by approximating the probability distribution of the objective function to the true probability distribution (i.e., maximum likelihood estimation). In addition, it is also possible to identify a viewpoint appropriate for evaluating the training menu included in the learning data 214 by learning.

[0057] The objective function 212 generated by the above learning can be said to represent the decision-making criteria of the expert. For example, the objective function 212 in which the weighting value for the perspective of "muscular hypertrophy" is greater than the weighting value for the perspective of "muscular endurance" indicates that the expert has created a training menu that takes into account both the perspectives of muscular endurance and muscular hypertrophy, but places more importance on muscular hypertrophy than on muscular endurance.

[0058] (Optimization calculation) By using the objective function 212, an evaluation value for evaluating the quality of a training menu can be calculated. Therefore, the generating unit 202 may generate a training menu that maximizes the evaluation value calculated using the objective function 212. Any method may be used to solve the optimization problem using the objective function, state data, and constraint conditions. For example, the generating unit 202 may generate an optimal training menu using an optimization solver from the objective function 212, state data 211, and constraint conditions. The generating unit 202 may use a general application program such as IBM ILOG CPLEX, Gurobi Optimizer, or SCIP as the optimization solver.

[0059] (Switching objective functions) In addition, in the training support device 2A according to the present exemplary embodiment, a plurality of objective functions 212 prepared in advance may be stored in the storage unit 21A or the like. In this case, the generation unit 202 may generate the training menu 213 by using an objective function corresponding to the subject who will be trained, among the plurality of objective functions 212 prepared in advance. According to this configuration, it is possible to use an objective function 212 that is more suitable for the subject, among the plurality of objective functions 212, and therefore, in addition to the effect of the training support device 2 according to the exemplary embodiment 1, an effect of being able to generate a training menu 213 that is more suitable for the subject can be obtained.

[0060] For example, a plurality of objective functions 212 corresponding to the purpose of training may be stored in the storage unit 21A etc. In this case, the generation unit 202 can generate a training menu 213 that matches the purpose of the subject by using the objective function 212 corresponding to the purpose of the subject's training.

[0061] (Generating training menus including background music) The training support device 2A can also generate a training menu including BGM (Back Ground Music). This will be described with reference to FIG. 7. FIG. 7 is a diagram showing an example of generating a training menu including BGM. The example of FIG. 7 differs from the example of FIG. 4 in that the state data includes music properties, the contents of the constraint conditions and the objective function, and that in the generated training menu, music that serves as BGM is associated with each training event.

[0062] The song properties are data indicating songs that can be used as background music and their characteristics. For example, the song properties shown in FIG. 7 indicate that a song called "song 1" can be used as background music, that the popularity of the song is 80, and the usage history of the song as background music. Note that the song properties are not limited to the example shown in FIG. 7, as long as they indicate songs that can be used as background music and their characteristics. For example, the song title, genre, release date, album name, artist name, song length, volume, melody, tempo, and beat may be used as song properties. In addition, for example, the degree to which the song is suitable for dancing, the live feeling, the degree to which it gives a positive impression, the strength of the feeling of energy given by the song, whether or not an electronic instrument is used, whether or not it is an instrumental (song without vocals), whether or not it is similar to speech, etc. may be used as song properties.

[0063] When generating a training menu including BGM, in addition to the constraints on the training, it is also possible to set constraints on the BGM. The constraints shown in FIG. 7 include a condition that a new song must be used at least once. By using such constraints, the BGM is determined so that a new song is always included in one training session. Note that the definition of a new song may be determined in advance, and for example, a song released within the past six months may be defined as a new song.

[0064] Furthermore, when generating a training menu that includes BGM, an objective function including a viewpoint for selecting BGM is used. The viewpoint may be anything related to BGM. For example, the objective function shown in FIG. 7 includes "compatibility between the event and BGM" and "popularity of BGM" as viewpoints for selecting BGM, in addition to "exercise intensity" which is a viewpoint for evaluating the training menu.

[0065] The above-mentioned objective function for generating a training menu including BGM can be generated by learning using the learning data 214 that indicates the training menu to be applied in the state related to the training and the music to be played during the training. For example, the learning data 214 that indicates the training menu and event properties performed by the expert, the BGM played by the expert during the execution of the training menu and the music properties of the BGM may be used. In this way, an objective function that indicates the decision-making criteria when the expert selects the BGM can be generated.

[0066] The "compatibility between an event and BGM" can be evaluated using, for example, the history of music usage. In other words, music that has been used frequently or many times as BGM in a certain training event can be evaluated as being compatible with that event. In this way, it is sufficient to determine in advance what properties are used to evaluate the viewpoints. This also applies to training-related viewpoints.

[0067] The learning unit 204 may automatically select a viewpoint such as "compatibility between the sport and BGM" using any feature selection technology. "Teaching Risk" is an example of a feature selection method in inverse reinforcement learning that the learning unit 204 can use. Feature selection using "Teaching Risk" involves assuming ideal parameters in an objective function, comparing them with the parameters in the learning process, and selecting features (i.e., viewpoints) that reduce the difference between the two parameters as important features.

[0068] Of course, the feature selection technique that the learning unit 204 can use is not limited to "Teaching Risk." The learning unit 204 can also select features using, for example, the method disclosed in Published PCT Publication PCT / JP2020 / 032848.

[0069] In the example of Fig. 7, a training menu showing training events for each day of the week and the music to be used as the background music for the training is generated from the above-mentioned state data, constraint conditions, and objective function. For example, the training menu of Fig. 7 shows that the first training event to be performed on Monday is "Event 2," and the background music during this training event is "Music 1."

[0070] The training support device 2A can also generate a training menu in which multiple pieces of music are associated with one training event. The training support device 2A can also generate a training menu in which one or multiple pieces of music are associated with multiple training events to be performed consecutively. When presenting multiple pieces of music to the subject, the training support device 2A may present the multiple pieces of music as one playlist.

[0071] As described above, in the training support device 2A according to this exemplary embodiment, the generation unit 202 uses the objective function 212 learned using the learning data 214 including information indicating music to be played during training to generate the training menu 213 including music to be played during training. Therefore, according to the training support device 2A according to this exemplary embodiment, in addition to the effects achieved by the training support device 2 according to the exemplary embodiment 1, an effect of being able to generate a more attractive training menu 213 can be obtained.

[0072] (Display of search keywords) Furthermore, the training support device 2A may present the subject with music pieces to be used as background music during training, as specified above, and allow the subject to make a final decision on the music piece. This will be described with reference to Fig. 8. Fig. 8 is a diagram showing an example of a display screen for a training menu and background music.

[0073] 8 shows two training events, "Event 2" and "Event 5," and the required times for each event as a training menu for Monday. This display screen example also shows "PL1" as a recommended playlist corresponding to "Event 2," and shows "PL3" as a recommended playlist corresponding to "Event 5." These events and playlists were identified by the generation unit 202 using the objective function 212.

[0074] The generating unit 202 may display the training events and recommended playlists from "Event 5" onwards, for example, in response to an operation of scrolling the display screen horizontally. The generating unit 202 may also display the training events and recommended playlists from Monday onwards in response to a predetermined operation. The generating unit 202 may also display each song included in the playlist, or each song included in the playlist may be displayed in response to an operation by the subject.

[0075] In the example display screen of Fig. 8, if the subject wishes to adopt the recommended playlist as is, he / she may perform an operation to that effect. Also, the subject may not adopt the recommended playlist and may select music to be used as background music by himself / herself. The "keywords" in the example display screen of Fig. 8 are displayed by the search unit 203 as an aid for the subject in selecting music.

[0076] The search unit 203 displays, as a search term for searching for music, a term indicating a music-related viewpoint among the viewpoints for evaluating a training menu, which are indicated in the objective function 212. By including the search unit 203, the training support device 2A can obtain an effect of easily searching for music matching the viewpoint, in addition to the effect obtained by the training support device 2 according to the exemplary embodiment 1.

[0077] Specifically, the display screen example of FIG. 8 shows two keywords, "compatibility with the event" and "popularity." These keywords are terms that indicate aspects related to music among the aspects shown in the objective function. When a keyword is selected, the search unit 203 searches for music that matches the keyword from among music that can be selected as background music, and displays the search results. If the desired music is included in the displayed search results, the subject can select it as background music for training.

[0078] In addition, the search unit 203 may select the above-mentioned viewpoint by utilizing the structure of the objective function. For example, when the weight of the viewpoint of "popularity" is large in the learned objective function, the search unit 203 may display "popularity" as a keyword.

[0079] For example, suppose that the keyword "compatibility with event" is selected when selecting background music for "event 2." In this case, the search unit 203 may refer to the music properties as shown in Fig. 7, identify a predetermined number of songs that are most frequently or frequently used as background music for "event 2," and display these songs as candidates for background music.

[0080] The training support device 2A may be configured to accept registration of favorite songs and playlists. The training support device 2A may also display songs and playlists registered as favorites together with recommended playlists. This makes it easier for the subject to set background music that suits his / her taste.

[0081] (Processing flow) The process (training support method) executed by the training support device 2A will be described with reference to Fig. 9. Fig. 9 is a flow diagram showing the flow of the process executed by the training support device 2A. Note that the following describes an example of generating a training menu including music played during training.

[0082] In S31, the data acquiring section 201 acquires state data 211 indicating a state related to the training performed by the subject. In addition, in S31, the data acquiring section 201 may also acquire constraint conditions for generating a training menu.

[0083] In S32, the generation unit 202 generates a training menu 213 according to the state indicated by the state data 211 acquired in S31. Specifically, the generation unit 202 generates the training menu 213 by optimization calculation using an objective function 212 generated by inverse reinforcement learning using learning data 214 indicating a training menu to be applied in the state related to the training. This training menu 213 includes music to be played during training.

[0084] In S33, generation unit 202 causes output unit 23A to display training menu 213 generated in S32 and music to be played when training menu 213 is executed. Furthermore, search unit 203 causes output unit 23A to display search terms for searching for music. As described above, the search terms displayed by search unit 203 are terms indicating a viewpoint related to music among the viewpoints for evaluating the training menu, which are indicated in objective function 212. Note that music may be displayed one by one, or multiple songs may be displayed together as one playlist (see FIG. 8).

[0085] In S34, the search unit 203 determines whether or not to execute a search. For example, the search unit 203 may determine to execute a search when it detects that an operation to select a displayed keyword has been performed. If the result of the determination in S34 is YES, the process proceeds to S35, and if the result of the determination in S34 is NO, the process proceeds to S36.

[0086] In S35, the search unit 203 searches for songs using the keyword selected by the subject from among the keywords displayed in S33, and displays and outputs the search results on the output unit 23A. Note that the search unit 203 may search for songs using the keyword input by the subject or the narrowing down conditions selected by the subject.

[0087] In S36, the search unit 203 judges whether or not a piece of music to be used as BGM has been selected. The piece of music to be selected may be the one displayed in S33 or the one displayed in S35. The selection of the piece of music may be accepted via the input unit 22A. If the judgment in S36 is YES, the process proceeds to S37, and if the judgment in S37 is NO, the process returns to S34.

[0088] In S37, the generation unit 202 determines the music selected in S as the background music to be played during training. This completes the training menu 213 including the music to be played during training, and the process in FIG. 9 ends.

[0089] [Modifications] The execution entity of each process described in the above exemplary embodiment is arbitrary and is not limited to the above example. For example, a training support system having the same function as the training support device 2A can be constructed by a plurality of devices that can communicate with each other. For example, a training support system having the same function as the training support device 2A can be constructed by distributing each block shown in FIG. 5 among a plurality of devices.

[0090] [Software implementation example] Some or all of the functions of the learning device 1 and the training support devices 2 and 2A may be realized by hardware such as an integrated circuit (IC chip), or may be realized by software.

[0091] In the latter case, the learning device 1 and the training support devices 2, 2A are realized, for example, by a computer that executes instructions of a program, which is software that realizes each function. An example of such a computer (hereinafter, referred to as computer C) is shown in FIG. 10. The computer C includes at least one processor C1 and at least one memory C2. The memory C2 stores a program P for operating the computer C as the learning device 1 and the training support devices 2, 2A. In the computer C, the processor C1 reads and executes the program P from the memory C2, thereby realizing each function of the learning device 1 and the training support devices 2, 2A.

[0092] The processor C1 may be, for example, a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof. The memory C2 may be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), or a combination thereof.

[0093] The computer C may further include a RAM (Random Access Memory) for expanding the program P during execution and for temporarily storing various data. The computer C may further include a communication interface for transmitting and receiving data to and from other devices. The computer C may further include an input / output interface for connecting input / output devices such as a keyboard, a mouse, a display, and a printer.

[0094] Furthermore, the program P can be recorded on a non-transitory tangible recording medium M that can be read by the computer C. Such a recording medium M can be, for example, a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer C can obtain the program P via such a recording medium M. Furthermore, the program P can be transmitted via a transmission medium. Such a transmission medium can be, for example, a communication network or broadcast waves. The computer C can also obtain the program P via such a transmission medium.

[0095] [Additional Note 1] The present invention is not limited to the above-described embodiment, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means disclosed in the above-described embodiment are also included in the technical scope of the present invention.

[0096] [Additional Note 2] Some or all of the above-described embodiments can be described as follows. However, the present invention is not limited to the following described aspects.

[0097] (Appendix 1) A training support device comprising: a data acquisition means for acquiring status data indicating a status related to training performed by a subject; and a generation means for generating a training menu corresponding to the status indicated by the status data by performing an optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status related to the training.

[0098] (Appendix 2) The training support device according to claim 1, wherein the data acquisition means acquires constraint conditions when generating a training menu for the subject, and the generation means generates the training menu that satisfies the constraint conditions.

[0099] (Appendix 3) 3. The training support device according to claim 1, wherein the generation means generates the training menu by using an objective function suited to the subject from among a plurality of objective functions prepared in advance.

[0100] (Appendix 4) The training support device of any one of appendices 1 to 3, wherein the learning data includes information indicating music to be played during training, and the generation means generates the training menu including the music to be played during training.

[0101] (Appendix 5) The training support device according to claim 4, further comprising a search means for displaying, as a search term for searching for music, a term indicating a viewpoint related to music among the viewpoints for evaluating a training menu indicated in the objective function.

[0102] (Appendix 6) At least one processor A training support method comprising: acquiring status data indicating a status related to training performed by a subject; and generating a training menu corresponding to the status indicated by the status data by performing an optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status related to the training.

[0103] (Appendix 7) A training support program that causes a computer to function as a data acquisition means for acquiring status data indicating a status related to training performed by a subject, and a generation means for generating a training menu corresponding to the status indicated by the status data by performing optimization calculations using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in that status.

[0104] (Appendix 8) A learning device comprising: a data acquisition means for acquiring learning data indicating a training menu to be applied in a state related to training, and a learning means for generating an objective function for generating a training menu in accordance with the state by performing inverse reinforcement learning using the learning data.

[0105] (Appendix 9) A learning method including: at least one processor acquiring learning data indicating a training menu to be applied in a training state according to the state; and generating an objective function for generating a training menu according to the state by performing inverse reinforcement learning using the learning data.

[0106] (Appendix 10) Computer, A data acquisition means for acquiring learning data indicating a training menu to be applied in a training state according to the state, and a learning program for causing the data acquisition means to function as a learning means for generating an objective function for generating a training menu according to the state by performing inverse reinforcement learning using the learning data.

[0107] [Additional Note 3] A part or all of the above-described embodiments can be further expressed as follows.

[0108] A training support device comprising at least one processor, the processor executing a data acquisition process for acquiring status data indicating a status related to training performed by a subject, and a generation process for generating a training menu corresponding to the status indicated by the status data by performing an optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status related to the training, by performing an optimization calculation using the objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the status.

[0109] The training support device may further include a memory that stores a program for causing the processor to execute the data acquisition process and the data generation process. The program may be recorded in a computer-readable, non-transitory, tangible recording medium.

[0110] A learning device comprising at least one processor, the processor executing a data acquisition process for acquiring learning data indicating a training menu to be applied in a training state according to the state, and a learning process for generating an objective function for generating a training menu according to the state by performing inverse reinforcement learning using the learning data.

[0111] The learning device may further include a memory, and the memory may store a program for causing the processor to execute the data acquisition process and the learning process. The program may be recorded in a computer-readable, non-transitory, tangible recording medium. [Explanation of symbols]

[0112] 1 Learning device 2. 2A Training Support Device 11, 21, 201 Data Acquisition Section 12, 204 Learning Department 22, 202 generation section 203 Search Department

Claims

1. A data acquisition means for acquiring status data indicating a status related to training performed by a subject; and a generation means for performing optimization calculations using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in a state related to training, thereby generating a training menu in accordance with the state indicated by the state data.

2. The data acquisition means acquires constraints for generating a training menu for the subject, The training support device according to claim 1 , wherein the generating means generates the training menu that satisfies the constraint condition.

3. The training support device according to claim 1 , wherein the generating means generates the training menu by using an objective function suited to the subject among a plurality of objective functions prepared in advance.

4. The learning data includes information indicating music to be played during training, The training support device according to claim 1 , wherein the generating means generates the training menu including music to be played during training.

5. 5. The training support device according to claim 4, further comprising a search means for displaying a term indicating a viewpoint related to music among viewpoints for evaluating a training menu indicated in the objective function as a search term for searching for music.

6. At least one processor Obtaining status data indicating a status related to training performed by a subject; generating a training menu corresponding to a state indicated by the state data by performing an optimization calculation using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the state related to the training.

7. Computer, A data acquisition means for acquiring status data indicating a status related to training performed by the subject; and A training support program that functions as a generation means for generating a training menu according to a state indicated by state data by performing optimization calculations using an objective function generated by inverse reinforcement learning using learning data indicating a training menu to be applied in the state related to the training.

8. A data acquisition means for acquiring learning data indicating a training menu to be applied in a training state according to the training state; and a learning means for generating an objective function for generating a training menu according to a state by performing inverse reinforcement learning using the learning data.

9. At least one processor Acquiring learning data indicating a training menu to be applied in a training state according to the training state; generating an objective function for generating a training menu according to a state by performing inverse reinforcement learning using the learning data.

10. Computer, A data acquisition means for acquiring learning data indicating a training menu to be applied in a training state according to the training state; and A learning program that functions as a learning means for generating an objective function for generating a training menu according to a state by performing inverse reinforcement learning using the learning data.

Images