Information processing device, optimization method, and control program

Abstract

An information processing device (1) executes optimization processing using an objective function that indicates the evaluation value of a solution candidate. The information processing device comprises an optimization processing unit (12) that executes optimization processing on the basis of a user's non-preference information.

Description

Information processing device, optimization method, and control program

[0001] The present disclosure relates to an information processing device, an optimization method, and a control program.

[0002] In recent years, a technique has been proposed for determining the content of a service to be provided to a service user in consideration of the user's preferences, interests, etc. (for example, see Patent Document 1).

[0003] The technology described in Patent Document 1 accepts specification of conditions for the service to be provided from the user, and based on the specified conditions, searches for and proposes from among a plurality of candidates those that are closest to the user's preferences.

[0004] Japanese Patent Application Laid-Open No. 2022-96954

[0005] The technology described in Patent Document 1 can identify and propose from among multiple candidates those that are closest to the user's preferences based on specified conditions. However, the user must change the conditions and receive new proposals until a satisfactory proposal is made, and if the user is unable to specify appropriate conditions, the number of repetitions of the series of processes increases. In other words, the technology described in Patent Document 1 has the problem of being unable to efficiently search for a solution.

[0006] The present disclosure has been made in view of the above, and aims to provide an information processing device that can reduce the number of interactions required to obtain a solution.

[0007] In order to solve the above-mentioned problems and achieve the objectives, the present disclosure is characterized by an information processing device that performs optimization processing using an objective function that indicates an evaluation value of a candidate solution, and that includes an optimization processing unit that performs optimization processing based on a user's non-preference information.

[0008] The information processing device according to the present disclosure has an effect of reducing the number of interactions required to obtain a solution.

[0009] FIG. 1 is a diagram showing a configuration example of an information processing device according to a first embodiment. FIG. 2 is a diagram showing an image of the overall operation of the information processing device according to the first embodiment. FIG. 3 is a flowchart showing an example of the operation of the information processing device according to the first embodiment. FIG. 4 is a diagram showing an image of the overall operation of the information processing device including the optimization processing device shown in FIG. 4. FIG. 5 is a diagram showing another image of the overall operation of the information processing device including the optimization processing device shown in FIG. 4. FIG. 1 is a diagram showing a configuration example of an optimization processing unit provided in an information processing device according to a second embodiment. FIG. 2 is a diagram showing another configuration example of an optimization processing unit provided in an information processing device according to a second embodiment. FIG. 3 is a diagram for explaining the optimization processing performed by an information processing device according to a third embodiment.

[0010] An information processing device, an optimization method, and a control program according to embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0011] First Embodiment. Fig. 1 is a diagram showing an example of the configuration of an information processing device 1 according to a first embodiment. The information processing device 1 derives an optimal solution that reflects a user's evaluation of solution candidates. The information processing device 1 performs optimization processing to derive the optimal solution based on the user's evaluation results of the solution candidates. Here, the user is a decision maker in the optimization processing, and the information processing device 1 performs optimization processing that reflects the will of the decision maker.

[0012] The information processing device 1 includes an information acquisition unit 11 that acquires information on the evaluation results of solution candidates from a user, an optimization processing unit 12 that executes optimization processing, a memory unit 13 that stores information on solution candidates to be presented to the user, and an output unit 14 that outputs the optimal solution derived by the optimization processing.

[0013] 2 is a diagram illustrating an overall operation of the information processing device 1 according to the first embodiment. The information processing device 1 derives an optimal solution by repeating a series of operations, for example, presenting pairs of solution candidates to a user, obtaining evaluation results that are the results of evaluations by a decision maker of the presented pairs of solution candidates, and performing an optimization process based on the obtained evaluation results to create new pairs of solution candidates and present them to the decision maker. Targets for which the information processing device 1 performs the optimization process to derive optimal solutions include, but are not limited to, meal menus, combinations of components to be installed in electronic devices, arrangements of components to be mounted on electronic device circuit boards, and layouts of equipment to be installed in facilities such as buildings.

[0014] 2, the information processing device 1 creates multiple pairs of candidate solutions to be compared and presents them to the decision maker, who then selects a more preferable candidate solution for each of the presented pairs of candidate solutions. The information processing device 1 may also allow the decision maker to select a less preferable candidate solution. In a configuration in which the decision maker selects a more preferable candidate solution, information about the candidate solution selected by the decision maker is preferred information, and information about the candidate solution not selected by the decision maker is non-preferred information.

[0015] The information processing device 1 acquires the selection results of the decision maker as evaluation results. The evaluation results include preference information and non-preference information. In the example shown in FIG. 2 , the information processing device 1 performs an optimization process using the preference information and non-preference information. The preference information and non-preference information are temporarily stored in a preference information database (DB) and a non-preference information database, respectively, and then used in the optimization process. Details of the optimization process will be described later. Note that the information processing device 1 may also perform an optimization process using at least one of non-preference information and preference information, for example, an optimization process using non-preference information. When performing an optimization process using non-preference information, the probability of creating pairs including solution candidates that do not match the decision maker's preferences is reduced, and pairs including solution candidates close to the optimal solution can be efficiently created, thereby shortening the time required to obtain an optimal solution. When having the decision maker select a more preferred solution candidate, it is expected that the number of solution candidates that are not selected will be greater than the number of solution candidates that are selected. In an optimization process using non-preference information that indicates candidate solutions that were not selected, it is possible to prevent the search from proceeding in the direction of undesirable candidate solutions, and it is expected that the number of times the process that requires the decision maker to select candidate solutions will be executed will be further reduced compared to an optimization process using preference information.

[0016] FIG. 3 is a flowchart showing an example of the operation of the information processing device 1 according to the first embodiment.

[0017] First, the information processing device 1 acquires solution candidates (step S11). The information processing device 1 acquires multiple solution candidates. The solution candidates may be created by the information processing device 1 based on information prepared in advance. That is, the information processing device 1 may acquire multiple solution candidates created by another external device from the other device, or may create multiple solution candidates based on prepared information. The information processing device 1 may perform optimization using an objective function for the acquired solution candidates. For example, if the solution candidates are meal menus, the menus may be optimized using protein, fat, carbohydrates, etc. as objective functions. The solution candidates are optimized using Bayesian optimization, a genetic algorithm, or the like. Note that when multiple solution candidates created by another external device are acquired, the information acquisition unit 11 acquires the solution candidates.

[0018] Next, the information processing device 1 acquires the evaluation results of the decision maker (step S12). Specifically, the information processing device 1 performs an interactive operation in which it combines two of the multiple solution candidates acquired in step S11 to create multiple pairs of solution candidates, presents them to the decision maker, and accepts a selection of which of the multiple pairs is preferable for each of the multiple pairs, thereby acquiring evaluation results for each of the presented solution candidates. The information processing device 1 may also accept a selection of which of the pairs is least preferable. The optimization processing unit 12 creates the solution candidate pairs, and the output unit 14 presents the solution candidate pairs to the decision maker. The information acquisition unit 11 acquires the decision maker's evaluation results.

[0019] Next, the information processing device 1 performs an optimization process (step S13). Specifically, the optimization processing unit 12 performs the optimization process (described later) based on the evaluation result acquired in step S12 as a process for deriving an optimal solution to be presented to a decision maker.

[0020] Next, the information processing device 1 determines whether to end the repetition (step S14). For example, the information processing device 1 determines that the repetition has ended when the number of repetitions reaches a predetermined number. The information processing device 1 may also determine that the repetition has ended when an operation instructing the end of the repetition is received from the decision maker.

[0021] If the information processing device 1 does not want to end the repetition (step S14: No), it executes steps S12 and S13 again. On the other hand, if the repetition is to end (step S14: Yes), it outputs the optimal solution obtained by repeatedly executing steps S12 and S13, and ends the operation (step S15).

[0022] Next, the optimization process executed by the optimization processing unit 12 of the information processing device 1 will be described with reference to FIGS.

[0023] 4 is a diagram showing an example of the configuration of the optimization processing unit 12 included in the information processing device 1 according to the first embodiment. The optimization processing unit 12 includes a probability distribution creation unit 21 that creates a probability distribution of the preference information and non-preference information acquired by the information acquisition unit 11, a utility function creation unit 22 that creates a utility function by converting the probability distribution created by the probability distribution creation unit 21, an objective function update unit 23 that updates an objective function indicating an evaluation value of a solution candidate based on the utility function created by the utility function creation unit 22, and a candidate extraction unit 24 that creates pairs of solution candidates based on the objective function. The probability distribution creation unit 21, the utility function creation unit 22, and the objective function update unit 23 constitute an objective function determination unit 20.

[0024] FIG. 5 is a diagram illustrating an overall operation of the information processing device 1 including the optimization processing unit 12 shown in FIG. 4. The information processing device 1 including the optimization processing unit 12 shown in FIG. 4 creates a probability distribution of the preference information and non-preference information included in the evaluation results obtained from the decision maker, and converts the created probability distribution to create a utility function. The information processing device 1 then optimizes solution candidates based on the utility function. The utility function indicates the evaluation value of the solution candidate to be presented to the decision maker. Note that the information processing device 1 including the optimization processing unit 12 shown in FIG. 4 may also create a probability distribution of one of the preference information and non-preference information, for example, a probability distribution of non-preference information, as shown in FIG. 6. FIG. 6 is a diagram illustrating another overall operation of the information processing device 1 including the optimization processing unit 12 shown in FIG. 4.

[0025] Fig. 7 is a flowchart showing an example of the operation of the optimization processing unit 12 shown in Fig. 4. The optimization processing unit 12 optimizes the solution candidates to be presented to the decision maker in accordance with the flowchart shown in Fig. 7.

[0026] 4, the optimization process for candidate solutions by the optimization processor 12 begins with the probability distribution generator 21 generating a probability distribution based on the evaluation results obtained from the decision maker (step S21), and then the utility function generator 22 converts the probability distribution into a utility function (step S22). A specific example of the operations in steps S21 and S22 will be described with reference to FIG.

[0027] FIG. 8 is a diagram illustrating the operation of the optimization processing unit 12 shown in FIG. 4 . FIG. 8 illustrates an example in which a probability distribution is created and converted into a utility function based on the evaluation results of each of the staple food candidates when a meal menu includes three types of staple food: rice, noodles, and bread. In the diagram, rice is referred to as "rice," noodles as "noodles," and bread as "bread." Similar notations will be used in the following explanation. In the example shown in FIG. 8 , a decision maker selects which of multiple prepared menu pairs is preferred. As a result, the number of times the menu with rice as the staple food was selected (determined to be preferred) is 8, the number of times the menu with noodles as the staple food was selected is 5, and the number of times the menu with bread as the staple food was selected is 1. Meanwhile, the number of times the menu with rice as the staple food was not selected is 1, the number of times the menu with noodles as the staple food was not selected is 5, and the number of times the menu with bread as the staple food was not selected is 6. In this case, the preference information is rice = 8, noodles = 5, and bread = 1, and the non-preference information is rice = 1, noodles = 5, and bread = 6. Using this preference information and non-preference information, a probability distribution generating unit 21 generates probability distributions for each of the preference information and non-preference information, and a utility function generating unit 22 converts the probability distributions for each of the preference information and non-preference information into utility functions.

[0028] The probability distribution creation unit 21 creates a probability distribution for each of the preference information and non-preference information using Bayesian estimation. In Bayesian estimation, a prior distribution for a certain hypothesis is first set, and then the prior distribution is updated based on the likelihood of new data collected thereafter, thereby creating a posterior distribution, which is the final conclusion. Bayesian estimation is expressed by equation (1).

[0029]

[0030] In equation (1), p(θ i |D) is the posterior distribution, p(θ i ) is the prior distribution, p(D|θ i ) is the likelihood.

[0031] The probability distribution creation unit 21 creates a probability distribution based on equation (1). As an example, creation of the probability distribution of preference information shown in Fig. 8 will be described. The prior distribution p(θ) is expressed by equation (2).

[0032]

[0033] In the case of FIG. 8, the preference information is rice=8, noodles=5, and bread=1, and the data D is given by equation (3).

[0034]

[0035] In this case, the likelihood of rice, noodles, and bread is p(D|θ i ) becomes equation (4).

[0036]

[0037] The posterior distribution p(θ i |D) is calculated by the prior distribution p(θ) and the likelihood p(D|θ i ) and this posterior distribution p(θ i |D) is the probability distribution of preference information.

[0038]

[0039] The posterior distribution p(θ i |D) is equation (5), for example, the posterior distribution p(θ rice |D) is expressed by equation (6).

[0040]

[0041] The probability distribution creation unit 21 also creates the probability distribution of non-preference information in the same manner. As explained with reference to FIG. 3, the optimization processing unit 12 repeatedly executes the optimization process. Therefore, when creating the probability distribution included in the repeatedly executed optimization process, the probability distribution created in the previous optimization process is used as the above-mentioned prior distribution p(θ i )

[0042] The above is the method for creating the probability distribution of preference information for the main dish of the meal menu. The probability distribution creation unit 21 also creates probability distributions of preference information for menu elements other than the main dish, and further creates a probability distribution P j is created according to equation (7).

[0043]

[0044] For example, if the menu candidate elements are rice, meat dishes, and tea, and the posterior distributions of the preference information are 8 / 14, 7 / 10, and 2 / 5, respectively, then the probability distribution P j = (8 / 14) * (7 / 10) * (2 / 5) = 4 / 25.

[0045] The probability distribution generation unit 21 generates a probability distribution Q of non-preference information for the entire meal menu. j For example, if the posterior distribution of the non-preference information of rice, meat dishes, and tea is 1 / 4, 2 / 7, and 2 / 5, then the probability distribution Q j = (1 / 4) * (2 / 7) * (2 / 5) = 1 / 35.

[0046] The utility function creation unit 22 calculates the probability distribution P of the preference information created by the probability distribution creation unit 21. j and the probability distribution Q of non-preferred information j is converted into a utility function. A utility function is a concept in economics that numerically expresses an individual's preferences. The utility function creation unit 22 converts the probability distribution P of preference information based on prospect theory or the like. j and the probability distribution Q of non-preferred information j is converted into a utility function.

[0047] For example, the utility function creation unit 22 generates a probability distribution P j The utility function F p,j , and the probability distribution of non-preferred information Q j The utility function F of non-preference information is calculated according to equation (9). q,j Convert to.

[0048]

[0049] FIG. 9 shows the utility function F p,j 10 shows an example of the utility function F q,j FIG. 9 is a diagram showing an example of τ p = 0.01, τ p = 1, τ p = 100, the utility function F p,j , and FIG. 10 shows an example of τ q = 0.01, τ q = 1, τ q = 100, the utility function F q,j Here is an example of τ p and τ q By adjusting the set value, the slope of the utility function for preference information and non-preference information can be adjusted to suit the characteristics of the decision maker. For example, a utility function can be created that takes into account the characteristics of the decision maker (gambler, conservative, etc.), such as the more preferences there are, the more they like it, and the more non-preferences there are, the more they dislike it.

[0050] The utility function creation unit 22 creates a utility function F using the preference information and non-preference information. r,j is created according to equation (10).

[0051]

[0052] For example, the probability distribution P j P j = 4 / 25, probability distribution of non-preferred information Q j GaQ j = 1 / 35, the utility function F r,j =F p,j +F q,j = 0.201 + 0.94 = 1.141.

[0053] Returning to the description of FIG. 7 , after the utility function creation unit 22 creates a utility function by converting the probability distribution in step S22, the objective function update unit 23 optimizes the solution candidates based on the utility function (step S23). For example, the objective function update unit 23 updates the evaluation value of the solution candidate by setting the utility function created by the utility function creation unit 22 as a new objective function. The objective function update unit 23 may set the utility function as a first objective function and further update the first objective function to a second objective function by optimizing the first objective function using Bayesian optimization, a genetic algorithm, or the like, thereby updating the evaluation value of the solution candidate. When updating the first objective function to the second objective function, a single second objective function may be created by weighting and adding multiple first objective functions, or multiple second objective functions may be created by individually updating each of the multiple first objective functions.

[0054] As described above, in the optimization processing unit 12 shown in FIG. 4, the objective function determination unit 20 creates a probability distribution of preference information and a probability distribution of non-preference information based on the preference information and non-preference information acquired by the information acquisition unit 11, and further converts the probability distribution to create a utility function, and optimizes candidate solutions based on the utility function.

[0055] In this embodiment, as a specific example, the creation of probability distributions and utility functions for staple foods (rice, noodles, bread) included in a meal menu has been described, but the information processing device 1 also creates and optimizes probability distributions and utility functions in a similar manner for main dishes, side dishes, and other items other than staple foods.

[0056] The candidate extraction unit 24 of the optimization processing unit 12 extracts a predetermined number of candidate solutions in descending order of evaluation value, and creates pairs of candidate solutions by combining two of the extracted candidate solutions. The candidate extraction unit 24 may also randomly extract candidate solutions from among candidate solutions with evaluation values ​​equal to or greater than a predetermined value to create pairs.

[0057] As described above, the information processing device 1 according to the first embodiment acquires evaluation results of solution candidates from a decision maker, and repeatedly executes an optimization process that updates an objective function indicating evaluation values ​​of the solution candidates based on the evaluation results of the decision maker, thereby deriving an optimal solution to be presented to the decision maker. Because the information processing device 1 according to the first embodiment performs an optimization process that reflects the evaluation results of the solution candidates by the decision maker, it is possible to reduce the number of interactions required to acquire the evaluation results of the solution candidates from the decision maker in order to derive an optimal solution.

[0058] Second Embodiment In the first embodiment, an information processing device 1 was described that derives an optimal solution by optimizing an objective function that indicates the evaluation values ​​of solution candidates based on the evaluation results of the decision maker. In contrast, an information processing device according to the second embodiment derives an optimal solution by optimizing the process of creating pairs of solution candidates to be presented to the decision maker. The configuration of the information processing device according to the second embodiment is the same as that of the information processing device 1 according to the first embodiment shown in FIG. 1. However, the content of the optimization process executed by the optimization processing unit 12 differs from that of the first embodiment. Therefore, in this embodiment, the optimization process that differs from that of the first embodiment will be described, and a description of the other parts will be omitted.

[0059] FIG. 11 is a diagram illustrating an example of an operation performed by the information processing device 1 according to the second embodiment, in which the process of creating pairs of solution candidates is optimized based on the evaluation results of the decision maker. As illustrated in FIG. 11 , the information processing device 1 optimizes the process of creating pairs of solution candidates based on preference information and non-preference information included in the evaluation results acquired from the decision maker. Note that the information processing device 1 may optimize the process of creating pairs of solution candidates based on either preference information or non-preference information. For example, as illustrated in FIG. 12 , the information processing device 1 may optimize the process of creating pairs of solution candidates based on non-preference information. FIG. 12 is a diagram illustrating another example of an operation performed by the information processing device 1 according to the second embodiment, in which the process of creating pairs of solution candidates is optimized based on the evaluation results of the decision maker.

[0060] The information processing device 1 that executes the operation shown in Fig. 11 updates, for example, an algorithm that extracts candidate solutions to be paired from a large number of prepared candidate solutions based on preference information and non-preference information. Fig. 13 shows a configuration example of an optimization processing unit 12a provided in the information processing device 1 that performs an optimization process to update the algorithm that extracts candidate solutions. Fig. 13 is a diagram showing a configuration example of the optimization processing unit 12a provided in the information processing device 1 according to the second embodiment.

[0061] The optimization processing unit 12a has a configuration in which the objective function determination unit 20 of the optimization processing unit 12 shown in FIG. 4 described in the first embodiment is replaced with an algorithm update unit 31. The algorithm update unit 31 updates an algorithm that extracts candidate solutions to be paired from a large number of candidate solutions based on the evaluation results of the candidate solutions by the decision maker. The algorithm update unit 31 adjusts parameters of the algorithm, for example, so that candidate solutions that are selected frequently by the decision maker are more likely to be extracted in the pair creation process, and so that candidate solutions that are selected less frequently by the decision maker are less likely to be extracted in the pair creation process.

[0062] The information processing device 1 may update the solution candidates based on the evaluation results of the decision maker. Fig. 14 shows a configuration example of an optimization processing unit 12b included in the information processing device 1 that updates the solution candidates through optimization processing. Fig. 14 is a diagram showing a configuration example of another optimization processing unit 12b included in the information processing device 1 according to the second embodiment.

[0063] The optimization processing unit 12b has a configuration in which the objective function determination unit 20 of the optimization processing unit 12 shown in FIG. 4 and described in the first embodiment is replaced with a candidate update unit 32. The candidate update unit 32 updates a large number of solution candidates that form the basis of pairs of solution candidates based on the decision maker's evaluation of the solution candidates. The candidate update unit 32, for example, adds new solution candidates that include elements constituting solution candidates that are frequently selected by the decision maker. For example, in the optimization process for deriving an optimal solution for a meal menu described in the first embodiment, if the number of times a menu item with rice as the staple food is selected is greater than the number of times other menu items with the staple food are selected, the candidate update unit 32 adds the menu item with rice as the staple food to the solution candidates. At this time, for example, if the number of times a menu item with bread as the staple food is selected is less than the number of times other menu items with the staple food are selected, a process of deleting some of the menu items with bread as the staple food from the solution candidates and updating the solution candidates may also be performed.

[0064] As described above, the information processing device 1 according to the second embodiment derives an optimal solution to be presented to the decision maker by repeatedly executing an optimization process that updates the process of creating pairs of solution candidates to be presented to the decision maker based on the evaluation results of the decision maker. As a result, similar to the information processing device 1 according to the first embodiment, the number of interactions required to obtain the evaluation results of the solution candidates from the decision maker can be reduced.

[0065] Third Embodiment In the first embodiment, an information processing device 1 was described that creates a probability distribution and converts it into a utility function for each of the preference information and non-preference information included in the evaluation result of the decision maker, and optimizes the objective function using Bayesian optimization or the like to derive an optimal solution. In the present embodiment, an information processing device is described that optimizes the objective function based on the preference information and non-preference information using a method different from that of the first embodiment. The configuration of the information processing device according to the third embodiment is the same as that of the information processing device 1 according to the first embodiment shown in FIG. 1. However, the content of the optimization process executed by the optimization processing unit 12 differs from that of the first embodiment. Therefore, in the present embodiment, the optimization process that differs from that of the first embodiment will be described, and a description of the other parts will be omitted.

[0066] FIG. 15 is a diagram for explaining the optimization process executed by the information processing device 1 according to the third embodiment. As shown in FIG. 15, the information processing device 1 according to the third embodiment converts non-preference information registered in the non-preference information DB to create pseudo-preference information equivalent to preference information. The information processing device 1 derives an optimal solution based on the preference information and non-preference information. The conversion of non-preference information into pseudo-preference information is performed by the optimization processing unit 12 of the information processing device 1. The configuration of the optimization processing unit 12 of the information processing device 1 according to the third embodiment is similar to the optimization processing unit 12 of the first embodiment shown in FIG.

[0067] 16 is a flowchart showing an example of the operation of the optimization processing unit 12 of the information processing device 1 according to the third embodiment. The optimization processing unit 12 of the information processing device 1 according to the third embodiment optimizes candidate solutions to be presented to a decision maker in accordance with the flowchart shown in FIG.

[0068] In the optimization operation of the solution candidates by the optimization processing unit 12 of the information processing device 1 according to the third embodiment, first, the probability distribution creation unit 21 converts non-preference information into pseudo-preference information (step S31). Specifically, the probability distribution creation unit 21 creates pseudo-preference information by converting the non-preference information into its reciprocal.

[0069] Next, the probability distribution creation unit 21 creates a probability distribution based on the preference information and the pseudo-preference information (step S32). Specifically, the probability distribution creation unit 21 adds the pseudo-preference information to the preference information, and creates a probability distribution for the preference information to which the pseudo-preference information has been added. The information processing device 1 creates the probability distribution in the same manner as in the first embodiment.

[0070] Next, the utility function creation unit 22 converts the probability distribution into a utility function (step S33). The utility function creation unit 22 converts the probability distribution into a utility function in the same manner as in the first embodiment. For example, the utility function creation unit 22 converts the probability distribution of the preference information to which the pseudo preference information has been added into a utility function F according to the above-mentioned equation (8). p,j The utility function generating unit 22 converts the utility function F of the non-preference information in the above-mentioned formula (9) into q,j F q,j = 1, and F r,j =F q,j+1 is output to the objective function update unit 23 as the utility function to be used in optimizing the solution candidates.

[0071] Next, the objective function update unit 23 optimizes the solution candidates based on the utility function input from the utility function creation unit 22 (step S34). The objective function update unit 23 optimizes the solution candidates in the same manner as in the first embodiment.

[0072] As described above, the information processing device 1 according to the third embodiment converts non-preference information included in the evaluation results of the solution candidates by the decision maker into pseudo-preference information equivalent to preference information, and optimizes the solution candidates based on the preference information and pseudo-preference information, thereby achieving the same effects as those of the information processing device 1 according to the first embodiment.

[0073] In this embodiment, an example has been described in which pseudo-preference information equivalent to preference information is created by converting non-preference information into its reciprocal. However, the information processing device 1 may also create pseudo-non-preference information equivalent to non-preference information by converting preference information into its reciprocal. That is, the information processing device 1 may create a utility function of non-preference information to which pseudo-non-preference information has been added based on the non-preference information and pseudo-non-preference information, and optimize solution candidates. In this case, the method of creating the utility function and the method of optimizing solution candidates are the same as when preference information and pseudo-preference information are used.

[0074] Fourth Embodiment Next, an information processing device according to a fourth embodiment will be described. The configuration of the information processing device according to this embodiment is the same as that of the first embodiment, but the information used in the optimization process is different from that of the first embodiment. In this embodiment, differences from the first embodiment will be described, and a description of common parts will be omitted.

[0075] 17 is a first diagram illustrating information used in the optimization process executed by the information processing device 1 according to the fourth embodiment. As shown in FIG. 17, the information processing device 1 according to the fourth embodiment classifies non-preference information into indifference information and dislike information, and executes the optimization process using the preference information, indifference information, and dislike information. The indifference information is information that is not considered important when a decision maker evaluates solution candidates and has little influence on the decision of which of a pair of solution candidates is preferable. The dislike information is information that has a large influence on the decision maker's decision of which of a pair of solution candidates is unpreferable when evaluating solution candidates.

[0076] The process of classifying non-preference information into indifference information and dislike information is performed, for example, by the probability distribution creation unit 21 of the optimization processing unit 12. The probability distribution creation unit 21 classifies the non-preference information registered in the non-preference information DB into indifference information and dislike information based on the results of comparing the preference information registered in the preference information DB with the non-preference information registered in the non-preference information DB. For example, if the solution candidates are meal menus and the number of times a menu with noodles as the staple food is selected as a preferred menu is approximately equal to the number of times it is selected as an unpreferred menu (the number of times it is not selected as a preferred menu), the probability distribution creation unit 21 classifies noodles as indifference information. Furthermore, for example, if the solution candidates are meal menus and the number of times a menu with bread as the staple food is selected as a preferred menu is low and the number of times it is selected as an unpreferred menu is high, the probability distribution creation unit 21 classifies bread as dislike information. Main dishes, side dishes, etc. are similarly classified into indifference information and dislike information.

[0077] After classifying the non-preference information into indifference information and dislike information, the optimization processing unit 12, for example, excludes the non-preference information corresponding to the indifference information and performs optimization processing using the non-preference information corresponding to the dislike information and the preference information. The optimization processing performed using the non-preference information corresponding to the dislike information and the preference information is the same as the optimization processing described in embodiment 1. In the optimization processing described in embodiment 2 or 3, the non-preference information corresponding to the dislike information and the preference information may also be used.

[0078] The optimization processing unit 12 may classify non-preference information into indifferent information and dislike information, and then create a probability distribution for each of the preference information, indifferent information, and dislike information using the method described in embodiment 1, and create a utility function to be used in optimizing candidate solutions based on the probability distribution for each of the preference information, indifferent information, and dislike information.

[0079] FIG. 18 is a second diagram illustrating information used in the optimization process executed by the information processing device 1 according to the fourth embodiment. As shown in FIG. 18 , the information processing device 1 may rank the preference information registered in the preference information DB and the non-preference information registered in the non-preference information DB. The ranking is performed for each element of the solution candidates. For example, if the solution candidates are meal menus, ranking is performed for each element, such as staple food, main dish, and side dish. For example, the information processing device 1 assigns a score N1 to elements of the solution candidates that are preferred by the decision maker, and a score N2 to elements of the solution candidates that are not preferred (N2<N1), with higher total scores being assigned to higher ranks. The information processing device 1 classifies information with lower ranks as dislike information and information with middle ranks as indifference information. After the classification is completed, the information processing device 1 performs the optimization process using the preference information, indifference information, and dislike information.

[0080] As described above, the information processing device 1 according to the fourth embodiment classifies non-preference information registered in the non-preference information DB into indifferent information and dislike information, and performs optimization processing using the preference information, indifferent information, and dislike information. This makes it possible to realize optimization processing that reflects the decision maker's preferences in more detail, and is expected to further reduce the number of interactions required to obtain evaluation results of solution candidates from the decision maker.

[0081] Fifth Embodiment Fig. 19 is a diagram showing a configuration example of an information processing device 1a according to a fifth embodiment. In Fig. 19, components common to the information processing device 1 according to the first embodiment shown in Fig. 1 are assigned the same reference numerals as in Fig. 1. In this embodiment, a description of components assigned the same reference numerals as in Fig. 1 will be omitted.

[0082] The information processing device 1a includes an information acquisition unit 11, an optimization processing unit 12, a storage unit 13, an output unit 14, and a preference proxy unit 15. The information processing device 1a has a configuration in which the preference proxy unit 15 is added to the information processing device 1 according to any one of the first to fourth embodiments.

[0083] The preference proxy unit 15 determines, on behalf of the decision maker, which of the pair of candidate solutions created by the optimization processing unit 12 is to be preferred. The preference proxy unit 15 determines which of the pair of candidate solutions is to be preferred based on preset preference conditions. For example, the preference proxy unit 15 determines which of the pair of candidate solutions is to be preferred using preference conditions 501 shown in FIG. 20 .

[0084] FIG. 20 is a diagram illustrating an example of preference conditions 501 used by the preference proxy unit 15 of the information processing device 1a according to the fifth embodiment. The preference conditions 501 are used by the preference proxy unit 15 when the solution candidate is a meal menu. The preference conditions 501 are composed of condition #1 and condition #2. Condition #1 indicates that a menu containing “rice,” “meat dish,” and “salad” has a higher priority than other menus, and a menu containing “bread,” “fish dish,” and “vinegared dish” has a lower priority than other menus. Condition #2 indicates that a menu containing “rice” has a higher priority than a menu containing “noodles,” which has a higher priority than a menu containing “bread,” and a menu containing “meat dish” has a higher priority than a menu containing “fish dish.” More detailed conditions, such as the type of meat dish, the type of fish dish, the type of rice, and the type of noodles, may be set. The information processing device 1a acquires and stores the preference conditions 501 from the decision maker in advance.

[0085] For example, if a pair of solution candidates consists of a first menu including "rice," a "meat dish," and a "salad" and a second menu that is not a menu that includes "rice," a "meat dish," and a "salad," the preference proxy unit 15 prefers (selects) the first menu in accordance with condition #1 of the preference conditions 501. Also, if a pair of solution candidates consists of a first menu including "bread," a "fish dish," and a "vinegared dish" and a second menu that is not a menu that includes "bread," a "fish dish," and a "vinegared dish," the preference proxy unit 15 prefers the second menu in accordance with condition #1 of the preference conditions 501. Also, if both the first menu and the second menu that make up the pair of solution candidates do not fall under the menus that include "rice," a "meat dish," and a "salad" or the menus that include "bread," a "fish dish," and a "vinegared dish," the preference proxy unit 15 determines the preferred menu based on condition #2 of the preference conditions 501.

[0086] As described above, the information processing device 1a according to the fifth embodiment is equipped with a preference proxy unit 15 that determines which of the candidate solutions to prefer on behalf of the decision maker, and therefore can perform the optimization process without interacting with the decision maker, thereby reducing the burden on the decision maker.

[0087] Sixth Embodiment Fig. 21 is a diagram showing a configuration example of an information processing device 1b according to a sixth embodiment. In Fig. 21, components common to the information processing device 1 according to the first embodiment shown in Fig. 1 are given the same reference numerals as in Fig. 1. In this embodiment, a description of components given the same reference numerals as in Fig. 1 will be omitted.

[0088] The information processing device 1b includes an information acquisition unit 11, an optimization processing unit 12, a storage unit 13, and an output unit 14, and the information acquisition unit 11 and the output unit 14 constitute a user interface unit 16 that interacts with a decision maker.

[0089] The output unit 14 of the user interface unit 16 presents the candidate solution pairs created by the optimization processing unit 12 to the decision maker by, for example, displaying them on a display device (not shown). The information acquisition unit 11 of the user interface unit 16 acquires the decision maker's preference results for the candidate solution pairs presented by the output unit 14. When the candidate solutions are meal menus, the output unit 14 of the user interface unit 16 displays, for example, the screen shown in FIG. 22 to present two menu plans to the decision maker. FIG. 22 is a diagram showing an example of a screen displayed by the output unit 14 of the user interface unit 16 included in the information processing device 1b according to the sixth embodiment. The screen shown in FIG. 22 includes a window 601 presenting menu plan 1, a window 602 presenting menu plan 2, and a group of operation buttons 610 for receiving input of the decision maker's evaluation results of menu plan 1 and menu plan 2.

[0090] As described above, the information processing device 1b according to the sixth embodiment is provided with a user interface unit 16 that interacts with the decision maker to obtain the evaluation results of the candidate solutions, and can efficiently obtain the decision maker's preference information and non-preference information.

[0091] Next, the hardware for realizing the information processing devices 1, 1a, and 1b described in each embodiment will be described. Since the information processing devices 1, 1a, and 1b are realized by similar hardware, the hardware for realizing the information processing device 1 according to the first embodiment will be described as an example.

[0092] 23 is a diagram illustrating an example of hardware for realizing the information processing device 1 according to the first embodiment. In FIG. 23, an example is shown in which the information processing device 1 is realized by a control circuit. The control circuit for realizing the information processing device 1 includes an input unit 91, a processor 92, a memory 93, and an output unit 94.

[0093] The input unit 91 receives signals from the outside. The output unit 94 outputs signals from the control circuit to the outside. The processor 92 is, for example, a CPU (Central Processing Unit), a central processing unit, a processing unit, an arithmetic unit, a microprocessor, a microcomputer, or a DSP (Digital Signal Processor). The memory 93 is, for example, a non-volatile or volatile semiconductor memory such as a RAM (Random Access Memory), a ROM (Read Only Memory), a flash memory, an EPROM (Erasable Programmable Read Only Memory), or an EEPROM (registered trademark) (Electrically Erasable Programmable Read Only Memory), a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a DVD (Digital Versatile Disk).

[0094] 23, a program for operating as the information acquisition unit 11 and optimization processing unit 12 of the information processing device 1 is stored in the memory 93, and the processor 92 reads and executes this program to realize the information acquisition unit 11 and optimization processing unit 12. The program stored in the memory 93 may be provided to a user or the like in a state written on a storage medium such as a CD (Compact Disc)-ROM or a DVD-ROM, or may be provided via a network.

[0095] The storage unit 13 and the output unit 14 of the information processing device 1 are realized by the memory 93 and the output unit 94 of the control circuit, respectively.

[0096] The information processing device 1 can also be realized by dedicated hardware, such as a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof.

[0097] The configurations described in the above embodiments are merely examples, and may be combined with other known technologies, or may be combined with other embodiments, or may include parts of the configurations that are omitted or modified without departing from the spirit of the invention. For example, while the embodiments have been described with reference to a case where the object of optimization is a meal menu, the object of optimization may also be the arrangement of components that make up an electronic device, the combination of components that make up an electronic device, or the like.

[0098] 1, 1a, 1b Information processing device, 11 Information acquisition unit, 12, 12a, 12b Optimization processing unit, 13 Memory unit, 14 Output unit, 15 Preference agent unit, 16 User interface unit, 20 Objective function determination unit, 21 Probability distribution creation unit, 22 Utility function creation unit, 23 Objective function update unit, 24 Candidate extraction unit, 31 Algorithm update unit, 32 Candidate update unit.

Claims

1. An information processing device that executes optimization processing using an objective function that indicates the evaluation value of a candidate solution, characterized by comprising: an optimization processing unit that executes the optimization processing based on a user's non-preference information.

2. The information processing device according to claim 1, further comprising an objective function determination unit that converts a first objective function indicating the evaluation value into a second objective function based on the non-preference information, and wherein the optimization processing unit performs optimization processing using the second objective function.

3. The information processing device according to claim 2, characterized in that the objective function determination unit converts the first objective function into the second objective function using a utility function created based on the non-preference information.

4. The information processing device according to claim 3, wherein the objective function determination unit creates the utility function by transforming the probability distribution of the non-preference information.

5. An information processing device as described in claim 4, further comprising an information acquisition unit that acquires the non-preference information, wherein when the information acquisition unit acquires second non-preference information after converting the probability distribution of first non-preference information to create a first utility function, the objective function determination unit updates the probability distribution based on the second non-preference information and updates the objective function based on a second utility function obtained by converting the updated probability distribution.

6. The information processing device according to claim 1, wherein the optimization processing unit determines candidate solutions to be presented to the user based on the objective function and the non-preference information.

7. The information processing device described in claim 6, characterized in that the optimization processing unit comprises: a candidate extraction unit that extracts a candidate solution to be presented to the user from a plurality of candidate solutions based on the objective function; and an algorithm update unit that updates the algorithm used by the candidate extraction unit in the extraction based on the non-preference information.

8. The information processing device described in claim 6, characterized in that the optimization processing unit comprises: a candidate extraction unit that extracts a candidate solution to be presented to the user from multiple candidate solutions based on the objective function; and a candidate update unit that updates the multiple candidate solutions to be extracted by the candidate extraction unit based on the non-preference information.

9. An information processing device according to any one of claims 1 to 8, characterized in that the optimization processing unit converts preference information indicating candidate solutions preferred by the user into an inverse number to create pseudo-non-preference information equivalent to the non-preference information, and executes the optimization process based on the non-preference information and the pseudo-non-preference information.

10. The information processing device according to claim 1, wherein the optimization processing unit executes optimization processing based on the user's preference information and the non-preference information.

11. An information processing device as described in claim 10, comprising: an information acquisition unit that acquires the preference information and the non-preference information; and an objective function determination unit that converts a first objective function indicating the evaluation value into a second objective function based on a first utility function created by converting the probability distribution of the preference information and a second utility function created by converting the probability distribution of the non-preference information, and determines the second utility function as the objective function to be used in the optimization process, wherein the objective function determination unit, after determining the objective function based on the first preference information and the first non-preference information, updates the objective function based on the second preference information or the second non-preference information acquired by the information acquisition unit if the information acquisition unit acquires one of the second preference information and the second non-preference information, and updates the objective function based on the second preference information and the second non-preference information acquired by the information acquisition unit if the information acquisition unit acquires both the second preference information and the second non-preference information.

12. The information processing device according to claim 10, wherein the optimization processing unit determines candidate solutions to be presented to the user based on the objective function, the non-preference information, and the preference information.

13. The information processing device described in claim 10 or 12, characterized in that the optimization processing unit converts the non-preference information into its reciprocal to create pseudo-preference information equivalent to the preference information, and performs the optimization processing based on the preference information and the pseudo-preference information.

14. An information processing device as described in any one of claims 10 to 13, characterized in that each of the candidate solutions includes a plurality of elements that are preferred by the user, and the optimization processing unit classifies each of the plurality of elements included in the candidate solutions that are not preferred by the user based on the user's preference results for the candidate solutions into elements that the user dislikes and elements that the user has little interest in, and uses information indicating the elements that the user dislikes as the non-preference information.

15. The information processing device described in claim 10, characterized in that each of the candidate solutions includes a plurality of elements that are preferred by the user, and the optimization processing unit classifies each of the plurality of elements included in the candidate solutions that are not preferred by the user into elements that the user dislikes and elements that the user has little interest in based on the user's preference results for the candidate solutions, and performs the optimization processing based on dislike information indicating the elements that the user dislikes, indifference information indicating elements that the user has little interest in, and the preference information.

16. An information processing device according to any one of claims 1 to 15, characterized in that it comprises a preference proxy unit that, when solution candidates are presented, selects the solution candidates required for the optimization process on behalf of the user based on the user's preference conditions for each of the presented solution candidates.

17. An information processing device according to any one of claims 1 to 16, further comprising a user interface unit that presents the solution candidates to the user and receives a preference for the presented solution candidates from the user.

18. An optimization method in which an information processing device executes an optimization process using an objective function that indicates an evaluation value of a candidate solution, the optimization method being characterized in that the optimization process is executed based on non-preference information of a user.

19. A control program for controlling an information processing device that executes an optimization process using an objective function that indicates an evaluation value of a candidate solution, the control program causing the information processing device to execute the optimization process based on a user's non-preference information.

Images