WO2018139300A1 - 情報処理装置、情報処理方法、及び、情報処理プログラムが記録された記録媒体 - Google Patents
情報処理装置、情報処理方法、及び、情報処理プログラムが記録された記録媒体 Download PDFInfo
- Publication number
- WO2018139300A1 WO2018139300A1 PCT/JP2018/001160 JP2018001160W WO2018139300A1 WO 2018139300 A1 WO2018139300 A1 WO 2018139300A1 JP 2018001160 W JP2018001160 W JP 2018001160W WO 2018139300 A1 WO2018139300 A1 WO 2018139300A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- information
- likelihood
- relationship
- state
- occurrence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Ceased
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
- G08G1/0116—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from roadside infrastructure, e.g. beacons
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N33/00—Investigating or analysing materials by specific methods not covered by groups G01N1/00 - G01N31/00
- G01N33/02—Food
- G01N33/025—Fruits or vegetables
Definitions
- the present invention relates to, for example, an information processing apparatus that predicts an event that occurs regarding a target.
- a parameter included in a mathematical model indicating the state related to the target is observed for the target.
- a simulation technique for simulating the target by adjusting it to match the information is expressed as probability distributions relating to the parameters, and the simulation is performed to obtain data so that values close to the observation information are calculated using a model including the parameters. There is a technique.
- FIG. 11 is a diagram conceptually showing processing in the simulation method.
- information in which the observation information 501 input to the observation model 502 and the system model 503 that is a mathematical simulation model are merged is created, and the state related to the object is estimated using the created information.
- the observation information 501 represents information observed regarding the object.
- the observation model 502 is a mathematical model that represents a state related to the object.
- the probability distribution relating to the variable (parameter) included in the observation model 502 or the system model 503 was observed for the object.
- a probability distribution that matches the observation information (observed value) 501 is obtained.
- state estimation 504 for obtaining a value of an objective variable representing a state of interest with respect to the target is executed.
- the prediction information (prediction value) and the estimation information (estimation value) obtained for the objective variable are represented as “analysis information”.
- filtering when time-series data is input, processing for calculating the influence of the observation information 501 on the analysis information is executed in the order of timing in the time-series data. This process is expressed as “filtering”. Examples of filters in the filtering include a particle filter and an ensemble Kalman filter.
- the probability distribution of the variable values before executing the filtering process is represented as “prior probability distribution” (or “prior distribution”).
- the probability distribution of variable values after executing the filtering process is represented as “posterior probability distribution” (or “posterior distribution”).
- Patent Documents 1 and 2 disclose a simulation technique for estimating a state related to an object by obtaining a likely state while using filtering.
- a method for obtaining a likely state is also called a maximum likelihood estimation method.
- Patent Literature 1 calculates the likelihood of the observation information based on the error between the observation information observed for the target and the prior probability distribution for the target in the process of obtaining the posterior probability distribution in filtering. .
- the apparatus creates the posterior probability distribution by applying weighting according to the calculated likelihood to the prior probability distribution.
- the apparatus disclosed in Patent Document 2 includes a plurality of observation models related to an object and a determination unit.
- the determination unit selects a plausible result based on a plurality of posterior distributions in the plurality of observation models in a data assimilation process using filtering.
- the apparatus is capable of highly accurately simulating a state related to an object even in non-ideal observation information (for example, data including noise).
- Patent Document 1 or Patent Document 2 Even if the apparatus disclosed in Patent Document 1 or Patent Document 2 is used, prediction regarding the object cannot be executed with high accuracy. In other words, even if these devices are used, it is not possible to solve the problem relating to an error generated between information (or a value) actually observed with respect to an object and analysis information predicted with respect to the object. The reason for this is that even if these devices determine the state related to the object by selecting a plausible state, there will be a discrepancy between the determined state and the state that actually occurs for the object. .
- one of the objects of the present invention is to provide an information processing apparatus or the like that provides information that is a basis for estimating a state related to a target with high accuracy.
- an information processing apparatus includes: A relationship determining means for determining a relationship between state information representing a state related to a target and the likelihood of occurrence of a scenario representing a mode in which the state information changes; Evaluation processing means for obtaining the state information when the probability of occurrence satisfies a predetermined selection condition based on the relationship determined by the relationship determination means.
- an information processing method includes: The calculation processing device determines a relationship between the state information indicating the state related to the object and the likelihood of occurrence of a scenario indicating a mode in which the state information changes, and the probability of occurrence is predetermined based on the determined relationship. The state information when the selection condition is satisfied is obtained.
- an information processing program A relationship determination function for determining a relationship between state information representing a state related to an object and the likelihood of occurrence of a scenario representing a mode in which the state information changes; Based on the relationship determined by the relationship determination function, the computer realizes an evaluation processing function for obtaining the state information when the likelihood of occurrence satisfies a predetermined selection condition.
- this object is also realized by a computer-readable recording medium that records the program.
- the information processing apparatus and the like it is possible to provide information that is a basis for estimating a state related to a target with high accuracy.
- Data assimilation technology introduces information representing uncertainty into a mathematical model that represents the state of a target that changes over time so that it matches the observation information about the target. It is a technology that realizes higher prediction accuracy.
- the model in the data assimilation technique includes a plurality of parameters and state variables. A value is set for each of the plurality of parameters. Further, the value of the state variable at each timing is set in the state variable.
- state value or state information state in which the value of the parameter and the value of the state variable (hereinafter also referred to as state value or state information) change as time passes will be referred to as “scenario”. That is, the scenario represents an aspect in which the state related to the object changes.
- the data assimilation technique it is assumed that the values of the state variables are distributed according to the probability distribution. As a result, there are a plurality of scenarios estimated by the model.
- a plurality of particles represents a simulation (or the scenario itself) according to the plurality of scenarios.
- the values of a plurality of parameters at a certain timing and the values of the state variables are expressed as “the state at the certain timing”.
- the state that may actually occur with respect to the target is predicted probabilistically.
- the proportion of particles related to a scenario that satisfies the condition that the state at timing t is state C is 80% of all particles, the probability that the state at timing t is state C is high. Expected to be 80%. There may be a plurality of particles related to one scenario.
- the analysis information represents the value of the state variable obtained based on the mathematical model related to the object.
- the observation information represents, for example, observation information (observation value) observed by an observation apparatus that observes the target.
- the model value represents a value related to a variable representing a state in one scenario in a model in which values of variables included in the mathematical model are distributed according to a probability distribution.
- particles are duplicated or deleted so as to obtain a probability distribution in which observation information at a certain timing is created for a plurality of model values at a certain timing.
- a scenario regarding particles is changed so that a plurality of model values at a certain timing has a probability distribution in which observation information at the certain timing is created.
- filtering process the process of copying, deleting, or changing the particles.
- a technique for realizing the filtering process there are a technique such as a particle filter and an ensemble Kalman filter.
- a particle that is most suitable (applicable and likely) to observation information at one timing is selected from among a plurality of particles.
- the degree of fit is expressed as “likelihood”.
- the first likelihood (likelihood for each observation information related to a certain timing) is observed by a j-th (j is a natural number) observation device at a certain timing t (t is a natural number) as shown in Equation 1.
- j is a natural number
- t is a natural number
- Equation 1 Represents the degree to which the variable x t
- B) represents the conditional probability that event A occurs when event B occurs.
- t ⁇ 1 (i) represents the state estimated for the timing t based on the state at the timing (t ⁇ 1) and the mathematical model for the i-th particle.
- the first likelihood represents the degree to which the analysis information is applied to the observation information y j, t at the timing t in the i-th scenario.
- the second likelihood (the total likelihood for a certain timing) is the variable x t
- j j represents multiplication for j.
- the second likelihood represents the degree to which analysis information is applied to all observation information at timing t in the i-th scenario.
- the third likelihood (likelihood for each observation information regarding a certain period) is the first likelihood (Equation 1) in the period (t_1 ⁇ t_2) from timing t1 to timing t2. Represents the power of.
- the third likelihood represents the degree to which the analysis information is applied to the observation information y j, t in the period from the timing t1 to the timing t2 in the i-th scenario.
- the fourth likelihood (total likelihood for a certain period) is the second likelihood (Expression 2) in the period (t1 ⁇ t2) from timing t1 to timing t2. Represents the square.
- the fourth likelihood represents the extent to which analysis information is applied to all observation information in the period from timing t1 to timing t2 in the i-th scenario.
- the third likelihood and the fourth likelihood do not necessarily have to be a processing procedure that is calculated by raising to the power, as exemplified in Equations 3 and 4, for example, by calculating the sum. It may be a processing procedure.
- the processing procedure for calculating the third likelihood and the fourth likelihood is not limited to Expression 3 and Expression 4.
- the particle average likelihood is calculated based on the average values for all scenarios with respect to the four likelihoods of the first likelihood to the fourth likelihood described above.
- ⁇ represents a process for calculating the sum.
- the first average likelihood shown in Expression 5 represents an average value obtained by averaging the degree to which the analysis information is applied to the observation information y j, t at the timing t with respect to all scenarios.
- the second average likelihood shown in Expression 6 represents an average value obtained by averaging the degree to which the analysis information is applied to all the observation information at the timing t with respect to all scenarios.
- the third average likelihood shown in Expression 7 represents an average value obtained by averaging the degree to which the analysis information is applied to the observation information y j, t in the period from the timing t1 to the timing t2, for all scenarios.
- the fourth average likelihood shown in Expression 8 represents an average value obtained by averaging the degree to which the analysis information is applied to all the observation information in the period from the timing t1 to the timing t2 with respect to all the scenarios.
- particles having a high likelihood at one timing are selected from among a plurality of particles.
- the simulation according to the scenario related to the particle having the highest likelihood at one timing does not necessarily have high prediction accuracy related to the object.
- the inventor of the present application has found that the uncertainty in the model related to the simulation is expressed by using a finite discrete distribution, and that there is one reason why the prediction accuracy related to the target does not increase.
- the inventor of the present application has a statistical relationship between an evaluation value such as likelihood related to particles (described later with reference to Equation 9) and analysis information, and particles out of the relationship are It has been found that even if a particle has a high likelihood, the state estimated according to the scenario regarding the particle is in a state of being deviated from the actual observation information. Therefore, the present inventor has found a problem that prediction accuracy using the simulation is low even in a simulation according to a scenario relating to particles having high likelihood.
- FIG. 1 is a block diagram showing the configuration of the information processing apparatus 101 according to the first embodiment of the present invention.
- the information processing apparatus 101 roughly includes a simulation unit 102 and an evaluation processing unit 103.
- the simulation unit 102 includes a data acquisition unit 104, an assimilation calculation unit 105, a prediction unit 106, and a calculation end determination unit 107.
- the evaluation processing unit 103 includes an evaluation value calculation unit 108, a relationship determination unit 109, and a state calculation unit 110.
- the observation device 151 observes the state related to the object, creates observation information (observation value) representing the observed state, and stores the created observation information in the observation information storage unit 154.
- the input device 152 inputs setting information from the outside, and stores the input setting information in the setting information storage unit 155.
- the setting information includes information indicating a setting model related to the target, and an end condition (stopping criteria) indicating whether or not to end the assimilation process.
- the end condition is, for example, a condition that the processing timing has reached a predetermined timing (that is, end timing).
- the end condition may be a condition that any one of the state values (state information) satisfies a predetermined condition (for example, the state value exceeds a threshold value).
- the state value represents a state related to the object.
- the state information is, for example, information representing a state of interest among states related to a target such as a target crop or an expressway, and information indicating a yield of the target crop described later, a time required until the congestion is resolved, and the like. It is.
- the information processing apparatus 101 is connected (or communicably connected) to the observation information storage unit 154, the setting information storage unit 155, and the output device 153.
- the information processing apparatus 101 can read the observation information stored in the observation information storage unit 154 and the setting information stored in the setting information storage unit 155.
- the information processing apparatus 101 may output the processed result to the output apparatus 153.
- the data acquisition unit 104 reads the observation information stored in the observation information storage unit 154 and outputs the read observation information to the assimilation calculation unit 105.
- the assimilation calculation unit 105 inputs the observation information output from the data acquisition unit 104. Next, the assimilation calculation unit 105 creates a plurality of scenarios (particles). This process will be described.
- the assimilation calculation unit 105 creates a plurality of sets including state information (state values) at a certain timing. That is, the assimilation calculation unit 105 creates a plurality of scenarios (particles) including the state information (state values). Next, the assimilation calculation unit 105 calculates a random number having a certain variance around the observation information, and creates a plurality of sets configured by combining the created random numbers. The assimilation calculation unit 105 creates a scenario (particle) including each set. Therefore, the above processing is processing for copying, changing, or deleting particles so as to obtain a probability distribution in which observation information at a certain timing is created.
- the assimilation calculation unit 105 performs the filtering process based on the observation information regarding each timing and the mathematical model by the above-described processing.
- the assimilation calculation unit 105 is based on the input observation information, and is described above with reference to the likelihood (degree of likelihood) as described above with reference to Equations 1 to 4, and with reference to Equations 5 to 8.
- At least one likelihood is calculated from the average likelihood (the degree of the probability of occurrence averaged with respect to the scenario).
- the assimilation calculation unit 105 outputs the calculated likelihood and the average likelihood to the evaluation processing unit 103.
- the assimilation calculation unit 105 outputs the created particles to the prediction unit 106. With respect to the timing when the observation information is not observed, the assimilation calculation unit 105 outputs the particles to the prediction unit 106 without executing the filtering process described above.
- the prediction unit 106 receives the particles (scenario) output from the assimilation calculation unit 105, and the state information at the timing t in the input scenario and the differential equation representing the state in which the state related to the object changes with the transition of time are discrete. Based on the converted mathematical model, state information at timing (t + 1) is created.
- the calculation end determination unit 107 reads the end condition regarding whether to end the calculation from the setting information stored in the setting information storage unit 155, and determines whether or not to end the assimilation process based on the read end condition. Determine. When the calculation end determination unit 107 determines to end the assimilation process, the calculation end determination unit 107 ends the assimilation process. When the calculation end determination unit 107 determines not to end the assimilation process, the calculation end determination unit 107 executes the process shown in step S101 (described later with reference to FIG. 2).
- the simulation unit 102 evaluates the likelihood calculated in the process of executing the assimilation process (illustrated in Expressions 1 to 8) and the state information of each particle calculated as a result of executing the assimilation process. 103 for output. Therefore, the simulation part 102 estimates the scenario showing the aspect in which the state regarding an object changes by performing the process as mentioned above.
- the evaluation value calculation unit 108 inputs the state information output by the simulation unit 102 and the likelihood.
- the evaluation value calculation unit 108 executes a predetermined evaluation process on the input likelihood to thereby evaluate the degree of similarity between the particle (scenario) and the state actually generated in the target ( Equation 9) is calculated.
- ⁇ j that the total sum relating to j is calculated.
- ⁇ j represents a weight.
- ⁇ j ⁇ j 1.
- the predetermined evaluation process is, for example, a process for calculating an evaluation value according to Equation 9.
- the evaluation value calculation unit 108 outputs the calculated evaluation value to the relationship determination unit 109.
- the evaluation value calculation unit 108 performs the above-described processing for each particle.
- the relationship determination unit 109 inputs the evaluation value output by the evaluation value calculation unit 108 for each particle. As illustrated in FIG. 3 (described later), the relationship determining unit 109 is compatible with at least one value of state information in each particle and an evaluation value related to the particle. Ask for.
- the relationship determination unit 109 executes, for example, a fitting process for obtaining a relationship in which the value and the evaluation value match for a plurality of particles.
- the fitting process is a process for obtaining a coefficient in a predetermined relationship based on the value and the evaluation value.
- the predetermined relationship is, for example, a Gaussian function, a quadratic function, or a composite function obtained by adding these functions. When the predetermined relationship is an upward convex function, a value having a high evaluation value can be obtained by searching for a large value in the relationship.
- the relationship determination unit 109 may perform the above-described fitting process on the value when the state information satisfies a predetermined constraint condition. For example, in the case of simulation related to the process of growing a crop, the relationship determination unit 109 has a predetermined range of “50 to 150 tons per hectare” with respect to the crop yield x which is one of the state information.
- the fitting process as described above may be executed only for the state information that satisfies the constraint conditions. This is applicable, for example, when the amount of crops that can be harvested per hectare is obtained in advance.
- the relationship determining unit 109 outputs the calculated relationship to the state calculating unit 110. The fitting process will be described later with reference to FIG.
- the state calculation unit 110 receives the relationship output by the relationship determination unit 109, creates estimated information regarding the state information based on the input relationship, and outputs the created estimated information to the output device 153.
- the state calculation unit 110 obtains, as the estimation information, state information (for example, state information that is likely to be generated) when the ease of occurrence satisfies a predetermined selection condition in the input relationship.
- state information for example, state information that is likely to be generated
- the state calculation unit 110 has a large evaluation value indicating the likelihood of occurrence in the input relationship. Is obtained as the estimated information.
- the state calculation unit 110 obtains state information when the evaluation value is a large value by comparing the evaluation values included in the input relationship.
- the state calculation unit 110 may obtain the state information when the evaluation value in the relationship is maximum (or substantially maximum) as the estimation information.
- the substantially maximum may be a value within a predetermined range (for example, 3%, 5%, or 10%) from the maximum.
- FIG. 3 is a diagram conceptually showing the relationship subjected to the fitting process based on the state value related to the state in the scenario and the evaluation value related to the scenario.
- the horizontal axis of FIG. 3 represents the state value of interest in the state information, and the right side indicates that the state value is large, and the left side indicates that the state value is small.
- the vertical axis in FIG. 3 represents the evaluation value, and the higher the evaluation value is, the higher the evaluation value is (ie, the likelihood), and the lower the evaluation value, the lower the evaluation value (ie, the likelihood is not).
- a black dot represents a position determined by an evaluation value related to a scenario and a value of a state of interest among state information in the scenario.
- the curve represents the relationship determined based on the position determined for each particle. In the fitting process, a relationship with little deviation from the position determined for each particle is required.
- the process may be executed without using information (ie, outlier value) regarding a position that is different from the relationship.
- the process is a process for suppressing the influence of the outlier value on the parameter value in the relationship.
- the process for suppressing the influence of the outlier value may be, for example, a process that does not use a position greatly deviating from the distribution of the position of interest.
- the determination process of whether or not there is a divergence is performed by determining whether or not the position is more than “3 ⁇ ⁇ ” ( ⁇ represents a standard deviation) from the average position of all positions. Executed.
- the determination process of whether or not there is a divergence is executed, for example, by determining whether or not the position is within a predetermined distance from the average position of all positions.
- the process for suppressing the influence of the outlier value may be a process based on a predetermined constraint condition as described above.
- the distance represents a mathematical (mathematical) distance.
- the process of suppressing the influence of the outlier value may be a process of selecting an outlier value according to a cluster obtained according to a method of clustering a plurality of positions, for example.
- the clustering method is, for example, the K-means method or the X-means method.
- a cluster is obtained by clustering a plurality of positions, and the number of positions belonging to the obtained cluster is obtained.
- the obtained cluster is then classified into a cluster having a large number of belonging positions and a cluster having a small number of belonging positions, and the positions belonging to the latter cluster are outliers. Choose as.
- the estimated information can be obtained more accurately by the process of suppressing the influence of the outlier value.
- FIG. 2 is a flowchart showing the flow of processing in the information processing apparatus 101 according to the first embodiment.
- the data acquisition unit 104 determines whether observation information is stored in the observation information storage unit 154 (step S102). When observation information is stored in observation information storage unit 154 (YES in step S102), data acquisition unit 104 acquires observation information from observation information storage unit 154, and assimilation calculation unit 105 converts the observed observation information. Output for.
- the assimilation calculation unit 105 receives the observation information output from the data acquisition unit 104.
- the assimilation calculation unit 105 executes the assimilation process (step S103) and the filtering process as described above. In the assimilation process, the assimilation calculation unit 105 calculates at least one of the likelihoods as described above with reference to Equations 1 to 8, based on the obtained observation information.
- observation information is not stored in observation information storage unit 154 (NO in step S102)
- the assimilation process shown in step S103 is not executed.
- the data acquisition unit 104 outputs a signal indicating that there is no observation information to the prediction unit 106.
- the prediction unit 106 receives the state information output from the assimilation calculation unit 105 and the likelihood, or the signal output from the data acquisition unit 104.
- the prediction unit 106 creates state information at a timing next to the timing related to the input state information based on the input state information and a mathematical model (step S104).
- the calculation end determination unit 107 reads an end condition regarding whether or not to end the assimilation process from the setting information stored in the setting information storage unit 155, and determines whether or not to end the assimilation process based on the read end condition. Is determined (step S105). If calculation end determination unit 107 determines to end the assimilation process (YES in step S105), calculation end determination unit 107 ends the assimilation process. If the calculation end determination unit 107 determines not to end the assimilation process (NO in step S105), the calculation end determination unit 107 executes a process for advancing the timing (step S101). After the process shown in step S101, the process shown in step S102 is executed.
- the evaluation value calculation unit 108 calculates the third likelihood (illustrated in Equation 3) of each particle calculated by the assimilation calculation unit 105, the state information of each particle calculated by the prediction unit 106, and A predetermined evaluation process is executed. Through this process, the evaluation value calculation unit 108 calculates an evaluation value (illustrated in Expression 9) indicating the degree of similarity between the particles (scenario) and the state actually generated in the target (step S106).
- the relationship determination unit 109 matches the evaluation value (illustrated in Equation 9) calculated by the evaluation value calculation unit 108 for each particle and the state information calculated by the prediction unit 106 for the particle.
- a relationship is obtained (step S107).
- the relationship determination unit 109 obtains the relationship by executing a process related to fitting as illustrated in FIG.
- the evaluation value calculation unit 108 creates state information when the evaluation value is large based on the relationship obtained by the relationship determination unit 109 (step S108).
- the evaluation value calculation unit 108 creates state information when the evaluation value is the largest, for example, by obtaining state information when the Gaussian function is maximum.
- the information processing apparatus 101 can provide information that is a basis for estimating a state related to a target with high accuracy.
- the reason for this is that, as described above, the state that gives a likely state is determined based on the qualitative relationship between the evaluation value of the particle (scenario) about the object and the state information about the object. This is because it is possible to reduce the possibility that a prediction deviating from the observation information related to the object will be executed. The reason will be specifically described.
- the solution space represents one coordinate axis related to the harvest amount when the target is, for example, the crop yield.
- the yield in each particle is distributed on the coordinate axis (for example, on the coordinate axis indicated by “state value” in FIG. 3).
- the values distributed on the coordinate axes are not necessarily the states that actually occur in the object. It does not necessarily represent.
- the information processing apparatus 101 obtains the relationship (curve in FIG. 3) between the evaluation value calculated for each scenario and the state information indicating the state of interest in the scenario. In accordance with the relationship, plausible state information is obtained. In this case, for example, the information processing apparatus 101 determines the optimum parameter value based on the evaluation value and the state information (that is, the position of the black dot in FIG. 3). The correct estimation information.
- the predetermined relationship is continuous, for example, in a section where state information is distributed. As a result, there is a high possibility that the solution obtained by the information processing apparatus is a state that actually occurs in the object, compared to the case where discrete state information is processed.
- the information processing apparatus 101 obtains a relationship that matches the evaluation value and the state information, and obtains state information having a large evaluation value according to the obtained relationship. Therefore, it is not necessary to create an enormous number of particles for the purpose of providing information that is a basis for estimating the state of the object with high accuracy.
- the apparatus disclosed in Patent Document 1 or Patent Document 2 selects state information having a high evaluation value, the object is to provide information serving as a basis for estimating the state related to the object with high accuracy. For this purpose, it is necessary to create a huge number of particles.
- the process in the information processing apparatus 101 has been described with reference to the assimilation process as an example, but the process in the information processing apparatus 101 is not necessarily a process related to the assimilation process.
- the processing in the information processing apparatus 101 only needs to be performed regarding processing for predicting a state related to a target, and analyzing a dependency relationship between the predicted state and observation information observed for the target, and is limited to the above-described example.
- the subsequent embodiments are not limited to the assimilation process.
- FIG. 4 is a block diagram showing the configuration of the information processing apparatus 201 according to the second embodiment of the present invention.
- the information processing apparatus 201 roughly includes a plurality of simulation units 202 and an evaluation processing unit 203.
- the simulation unit 202 includes a data acquisition unit 204, an assimilation calculation unit 205, a prediction unit 206, a calculation end determination unit 207, and a statistical processing unit 208.
- the evaluation processing unit 203 includes an evaluation value calculation unit 209, a relationship determination unit 210, and a state calculation unit 211.
- the data acquisition unit 204 has the same function as that of the data acquisition unit 104.
- the assimilation calculation unit 205 has the same function as that of the assimilation calculation unit 105.
- the prediction unit 206 has the same function as that of the prediction unit 106.
- the calculation end determination unit 207 has the same function as the function that the calculation end determination unit 107 has.
- the evaluation value calculation unit 209 has the same function as that of the evaluation value calculation unit 108.
- the relationship determining unit 210 has a function similar to that of the relationship determining unit 109.
- the state calculation unit 211 has the same function as the function that the state calculation unit 110 has. Therefore, the simulation part 202 estimates the scenario showing the aspect in which the state regarding object changes.
- the observation device 151 observes the state related to the object, creates observation information (observation value) representing the observed state, and stores the created observation information in the observation information storage unit 154.
- the input device 152 inputs setting information from the outside, and stores the input setting information in the setting information storage unit 155.
- the setting information includes information indicating a setting model related to the object and an end condition indicating whether or not to end the assimilation process.
- the end condition is, for example, a condition that the processing timing has reached a predetermined timing (that is, end timing).
- the end condition is a condition that any state value in the state information satisfies a predetermined condition (for example, the state value exceeds a threshold value).
- the information processing apparatus 201 is connected (or communicably connected) to the observation information storage unit 154, the setting information storage unit 155, and the output device 153.
- the information processing apparatus 201 can read the observation information stored in the observation information storage unit 154 and the setting information stored in the setting information storage unit 155.
- the information processing device 201 may output the processed result to the output device 153.
- the simulation unit 202 reads observation information, setting information, and a random seed that is an initial condition for generating pseudo-random numbers. .
- the assimilation calculation unit 205 reads, for example, a random number seed from the random number seed information 212, and generates a pseudo random number based on the read random number seed.
- the assimilation calculation unit 205 may read a pseudo random number or a random number instead of the random number seed.
- pseudorandom numbers or random numbers are collectively referred to as “random numbers”. The random number is used when calculating an initial state value or when redistributing particles.
- the assimilation calculation unit 205 reads a plurality of random number seeds different for each simulation unit 202.
- the assimilation calculation unit 205 executes an assimilation process and a filtering process as described in the first embodiment.
- the assimilation calculation unit 205 calculates at least one of the likelihoods (the degree of likelihood and the degree of likelihood) as described with reference to Equations 1 to 8. calculate.
- the statistical processing unit 208 performs statistical processing on the likelihood and state information of each particle calculated by the assimilation calculation unit 205 (step S203), thereby selecting representative likelihood and representative state information. .
- the statistical processing unit 208 uses the representative state information x kl (where k is a natural number indicating a random number seed, and l (el) represents l-th state information), For example, a weighted average obtained by weighting the state information in the scenario represented by each particle with the fourth likelihood (Equation 4) related to the particle is calculated.
- x kl (i) represents the value of the l-th state (variable) calculated based on the pseudo-random number generated according to the k-th random number seed for the i-th particle.
- the statistical processing unit 208 calculates average state information according to the processing shown in Expression 10.
- the statistical processing unit 208 further relates to, for example, a scenario represented by each particle as a representative likelihood L kj regarding observation information (where k is a natural number indicating a random number seed, j represents j-th state information).
- a weighted average in which the likelihood is weighted by the fourth likelihood (Equation 4) related to the particle is calculated.
- L kj (i) represents the average likelihood regarding the j-th state (variable) calculated based on the pseudo-random number generated according to the k-th random number seed for the i-th particle.
- L kj (i) represents the likelihood calculated according to Equation 7.
- the statistical processing unit 208 calculates the average likelihood (that is, average likelihood) according to the processing shown in Expression 11.
- Each simulation unit 202 outputs the representative likelihood calculated according to the processing shown in Expression 10 and Expression 11 and the representative state information to the evaluation value calculation unit 209. Therefore, each simulation unit 202 calculates the above-described value for each random number type for which each state process is calculated, not for each particle.
- the relationship determination unit 210 and the state calculation unit 211 execute processing similar to the processing described in the first embodiment.
- the state calculation unit 211 outputs the created estimation information to the output device 153.
- FIG. 5 is a flowchart showing the flow of processing in the information processing apparatus 201 according to the second embodiment.
- Each simulation unit 202 reads the random number seed from the random seed information. Each simulation unit 202 executes the processing as described above with reference to step S201 and step S202 in parallel, pseudo-parallel, or sequentially. Step S202 represents the processing shown in steps S101 to S105 in FIG. The process executed in step S202 will be specifically described with reference to FIG.
- the data acquisition unit 204 determines whether observation information is stored in the observation information storage unit 154 (step S102).
- the assimilation calculation unit 205 executes the assimilation process as described above based on the observation information (step S103).
- the assimilation calculation unit 205 calculates at least one of the likelihoods as described with reference to Expressions 1 to 8 based on the observation information. If observation information is not stored in observation information storage unit 154 (NO in step S102), assimilation calculation unit 205 does not execute the process shown in step S103.
- prediction unit 206 When observation information is not stored in observation information storage unit 154 (NO in step S102), prediction unit 206 creates state information regarding a state at the next timing based on a mathematical model (step S104). . When observation information is stored in observation information storage unit 154 (YES in step S102), prediction unit 206 creates state information regarding the state at the next timing for the calculated particle (scenario) (step S102). S104).
- steps S101 to S105 (FIG. 2) is repeatedly executed when a predetermined end condition is not satisfied (NO in step S105).
- the simulation unit 202 executes the processing described with reference to Equation 10 and Equation 11, thereby representing the typical likelihood and Representative state information is created (step S203 in FIG. 5).
- the evaluation value calculation unit 209 After all the simulation units 202 create representative likelihood and representative state information, the evaluation value calculation unit 209 generates a random number type based on the representative likelihood and representative state information. Each evaluation value is calculated (step S204).
- the evaluation value is, for example, a value obtained by averaging representative likelihoods for each scenario.
- the relationship determination unit 210 executes a fitting process related to the evaluation value calculated for each random number type with respect to the particle (scenario) and the specified state information among the representative state information in the scenario (step S205).
- the fitting process is a process for obtaining a relationship by calculating a value of a parameter in a predetermined relationship such as a Gaussian function.
- the state calculation unit 211 creates, as estimated information, a state value when the evaluation value is maximum for the obtained relationship (step S206), and outputs the created estimated information to the output device 153.
- Equation 9 The number of random numbers that is a basis for calculating the evaluation value as shown may be adjusted.
- the relationship determining unit 210 calculates the degree of fitness as described above in the process for obtaining the relationship, and further determines whether to create particles based on the obtained degree of fitness. For example, when the fitness level is low, the relationship determination unit 210 further determines to create particles. When the relationship determining unit 210 determines to create particles, the simulation unit 202 further creates particles.
- the fitting process is executed based on the representative state information and the representative likelihood for each random number type.
- the representative state information and the representative likelihood are used.
- the fitting process may be executed without referring to.
- the relationship determination unit 210 executes the fitting process based on the likelihood regarding each particle calculated according to each random number type and the state information on the particle.
- the relationship determining unit 210 may calculate the likelihood and state information when the likelihood regarding each random number type is the highest as representative likelihood or representative state information.
- the observation information storage unit 154 the observation information observed by the observation device 151 such as a soil moisture sensor, a leaf area index sensor, a plant elongation sensor, a leaf nitrogen concentration sensor, etc. that is observed for the plant is displayed for each observation day. Assume that it is stored.
- the data acquisition unit 204 reads the observation information stored in the observation information storage unit 154.
- the period from the germination of the plant until the fruit of the plant ripens is 200 days.
- the state of the plant in the 200 days is simulated.
- the state of interest is the actual total weight on the 200th day.
- the end condition set in the setting information storage unit 155 is, for example, a condition that the timing is 200 days or more.
- the plant growth model is information representing the relationship between the observation information and information representing the state of the plant.
- the assimilation calculation unit 205 creates 1000 particles (scenarios) using, for example, a pseudorandom number generated according to a random number type.
- the assimilation calculation unit 205 performs the assimilation process (step S103 in FIG. 2) as described above in the first embodiment based on the state of the created particles, the plant growth model, and the observation information observed for the plant. ).
- the observation information regarding the timing is not stored in observation information storage unit 154 (NO in step S102)
- assimilation calculation unit 205 does not execute the assimilation process (step S103).
- the prediction unit 206 estimates the state information at the next timing related to the plant by applying the plant growth model to the state information on the particles calculated by the assimilation calculation unit 205 (step S104). That is, the prediction unit 206 creates state information at the next timing.
- step S105 When a predetermined end condition (in this example, the condition that the timing is 200 days or more) is not satisfied (NO in step S105), the processing in steps S101 to S105 (FIG. 2) is performed at the next timing. Executed with respect to.
- simulation unit 202 ends the assimilation process.
- the simulation unit 202 performs an assimilation process on the input data as described above, thereby allowing likelihoods for each of the 1000 particles (for example, soil moisture content, leaf area index, elongation, likelihood for leaf nitrogen concentration for plants). ( ⁇ j, 0 ⁇ 200 (i) ) and state information about the particles (for example, actual total weight x (i) ) are calculated.
- timing representing the germination timing is represented as “0”, and the timing representing 200 days later is represented as “200”.
- the evaluation value calculation unit 209 calculates the evaluation value of each particle according to Equation 12 (step S204 in FIG. 5).
- the relationship determining unit 210 determines parameters ( ⁇ , ⁇ ) in the predetermined relationship according to the predetermined relationship as shown in Expression 13.
- ⁇ represents an average value of x.
- ⁇ 2 represents the variance of x.
- ⁇ represents the circumference ratio.
- exp represents an exponential function related to the number of Napier.
- x represents a parameter (variable) relating to the total weight.
- E represents a parameter (variable) relating to the evaluation value E (i) shown in Expression 12.
- the relationship determination unit 210 executes the fitting process by calculating the parameters ( ⁇ , ⁇ ) in the relationship as shown in Expression 13 (step S205). Based on the relationship calculated by the simulation unit 202, the state calculation unit 211 obtains a state value (that is, ⁇ ) when the evaluation value in the relationship is the maximum (step S206).
- the average state value of interest in this example, the total weight of the plant fruit
- 130 (t / ha) the crop at the most likely particle
- 125 (t / ha) which is a state value in a particle having the highest likelihood
- the information processing apparatus 201 is based on the observation information about the plant and the plant growth model about the plant, and the total weight of the fruit on the 200th day after the plant germination is 129 tons per hectare. presume.
- the state calculation unit 211 outputs the created estimation information to an output device 153 such as a display.
- the observation information storage unit 154 stores observation information observed by an observation device 151 such as a traffic counter and a vehicle number sensor installed at the entrance / exit of each expressway.
- the observation information is information observed every 5 minutes, for example.
- the observation information is information representing, for example, the position where the traffic jam occurs, the vehicle flow velocity at each position, the vehicle density, the number of vehicles flowing into each highway, and the number of vehicles flowing out from the highway.
- the simulation is a simulation related to traffic jam prediction.
- a traffic jam in 3 hours from the present time is simulated.
- the time required to clear the traffic jam (hereinafter referred to as “solving time”) is estimated.
- the predetermined end condition is a condition that the timing has passed 3 hours from the present time.
- the specified state information is the time required until the congestion is resolved. Processing will be described with reference to FIGS. 2 and 5.
- the data acquisition unit 204 determines whether observation information is stored in the observation information storage unit 154 (step S102). When observation information is stored in observation information storage unit 154 (YES in step S102), data acquisition unit 204 reads the observation information from observation information storage unit 154, and sends the read observation information to assimilation calculation unit 205. Output. When observation information is not stored in observation information storage unit 154 (NO in step S102), data acquisition unit 204 outputs a signal requesting to create state information at the next timing to prediction unit 206. To do.
- the assimilation calculation unit 205 inputs the observation information output from the data acquisition unit 204.
- the assimilation calculation unit 205 creates 1000 particles (that is, a scenario) based on the input observation information and the traffic jam model, using, for example, a pseudo-random number generated according to a random number seed.
- the assimilation calculation unit 205 creates 1000 particles by executing assimilation processing based on the observation information for the created 1000 particles (step S103).
- the assimilation calculation unit 205 uses at least one of the likelihoods (the degree of likelihood and the degree of likelihood) as described with reference to Equations 1 to 8 based on the observation information.
- One likelihood is calculated.
- the prediction unit 206 predicts state information at the next timing based on the updated state information on the particles and the traffic jam model (step S104).
- the processing shown in steps S101 to S105 (FIG. 2) is executed at each timing during a period in which a predetermined end condition is not satisfied (NO in step S105).
- the simulation unit 202 performs the assimilation process. finish.
- the simulation unit 202 calculates the likelihood related to each of the 1000 particles and the state information related to the particles by executing the processing described above.
- the likelihood is observation information on the particle, for example, the position where the traffic jam occurs, the speed at which the vehicle moves at each position, the density of the vehicle, the number of vehicles flowing into each highway, or from each highway Likelihood ( ⁇ j, 0 ⁇ 3h (i) ) regarding information such as the number of vehicles that flow out.
- the state information is, for example, the time (x k (i) ) required until the congestion is eliminated at the k-th position, which is designated state information.
- timing indicating the present is expressed as “0”, and the timing indicating after 3 hours is expressed as “3h”.
- the evaluation value calculation unit 209 calculates an evaluation value related to the particle based on the likelihood calculated by the simulation unit 202 regarding each particle and the state information related to the particle (step S204). For example, when the error regarding the observation information such as the position where the traffic jam occurs, the flow velocity of the vehicle at each position, the density of the vehicle, the number of vehicles flowing into each highway, the number of vehicles flowing out from the highway, etc. is ⁇ j In addition, the evaluation value for each particle is calculated as shown in Equation 14.
- the relationship determination unit 109 calculates the values of parameters ( ⁇ , ⁇ ) included in the relationship based on the relationship shown in Expression 15. Ask.
- ⁇ represents an average value of x.
- ⁇ 2 represents the variance of x.
- ⁇ represents the circumference ratio.
- exp represents an exponential function related to the number of Napier.
- x represents a parameter (variable) relating to the time required until the congestion is resolved.
- E represents a parameter (variable) related to the evaluation value E (i) shown in Expression 14.
- the relationship determining unit 109 calculates the relationship between the observation information on the particle and the evaluation value related to the particle by obtaining the value of the parameter ( ⁇ , ⁇ ). In other words, the relationship determining unit 109 performs the fitting process by calculating the value of the parameter (step S205).
- the state calculating unit 211 Based on the relationship obtained by the relationship determining unit 109, the state calculating unit 211 obtains state information (for example, x value information) when the value of the relationship shown in Expression 15 is the maximum in the obtained relationship. Is calculated as estimated information (that is, estimated value ⁇ ) (step S206).
- state information regarding the k-th position having the highest likelihood among the times required to eliminate the traffic jam that occurred at the k-th (k is a natural number) position (the time required for the traffic jam to be eliminated in this example). Assume 2.3 hours. Further, it is assumed that the state information regarding the k-th particle having the next highest likelihood (in this example, the time required until the congestion is eliminated) is 1.8 hours.
- the apparatus disclosed in Patent Document 1 and the like calculates the state information related to 2.3 hours when the likelihood is the highest as estimated information.
- the information processing apparatus 201 it is possible to calculate the estimation information regarding the target more accurately. This is because even if a particle has a high likelihood, a process for excluding the case where the state information in the particle is unrealistic is possible by executing the fitting process.
- a scenario is created based on the random number type, and the fitting process is executed based on the state information in the created scenario and the likelihood related to the scenario.
- the information processing apparatus 201 may perform a process of adjusting the number of random number seeds in the fitting process, for example, according to the degree of dispersion between the state information and the likelihood. If the relationship calculated by the relationship determination unit 210 does not sufficiently represent the relationship between the state information and the likelihood, the assimilation calculation unit 205 further reads the random number seed. Thereafter, the processing shown in steps S202 to S205 in FIG. 5 is executed.
- the fitting process is executed based on the representative evaluation value and the representative state information. However, the fitting process is executed based on all the evaluation values and all the state information. May be. Further, as a procedure for selecting a representative value, a particle having the highest evaluation value may be selected.
- the prediction unit (the prediction unit 106 in FIG. 1 or the prediction unit 206 in FIG. 4) creates state information at the next timing, but it is not necessarily required to be at the next timing. Instead, state information at different timings may be created. That is, the timing that is a target to be calculated by the prediction unit is not limited to the above-described example.
- the information processing apparatus 201 can provide information that is a basis for estimating a state related to a target with high accuracy. This reason is the same as the reason described in the first embodiment.
- information processing apparatus 201 Furthermore, according to the information processing apparatus 201 according to the present embodiment, information that is a basis for estimating a state related to a target with high accuracy can be provided in a short time. This is because the assimilation processing is executed in parallel.
- the information processing apparatus 201 it is possible to stably provide information that is a basis for estimating a state related to a target with high accuracy. This is because estimation information with less variation can be created by creating estimation information based on typical state information and typical likelihood.
- FIG. 6 is a block diagram showing the configuration of the information processing apparatus 301 according to the third embodiment of the present invention.
- the information processing apparatus 301 includes a relationship determination unit 302 and an evaluation processing unit 303.
- the information processing apparatus 301 is connected to an estimation apparatus (simulation apparatus) 304 so that communication is possible.
- the estimation device 304 estimates a plurality of scenarios representing the manner in which the state related to the object changes.
- the estimation device 304 estimates a plurality of the states (or estimates the states probabilistically) according to the assimilation process described with reference to steps S101 to S105 (FIG. 2).
- the information processing apparatus 301 can output to the information processing apparatus 301 the likelihood of the scenario with respect to the observation information observed for the target and the scenario.
- the information processing apparatus 301 can be realized using, for example, the function of the simulation unit 102 in FIG. 1 or the function of the simulation unit 202 in FIG.
- FIG. 7 is a flowchart showing the flow of processing in the information processing apparatus 301 according to the third embodiment.
- the information processing apparatus 301 inputs the scenario estimated by the estimation apparatus 304 regarding the target and the observation information observed regarding the target.
- the relationship determining unit 302 determines the relationship between the state in the input scenario (that is, the state estimated with respect to the target) and the likelihood that the state will occur (step S301). In the process of determining the relationship, for example, the relationship determining unit 302 inputs a relationship including a parameter, and determines the parameter that matches the state in the scenario and the likelihood that the state will occur.
- the process for determining the relationship is the fitting process as described above with reference to FIG.
- the evaluation processing unit 303 obtains a state when the likelihood is high in the relationship determined by the relationship determining unit 302. In other words, the evaluation processing unit 303 obtains a predetermined selection condition that represents a condition for selecting a likely state based on the relationship determined by the relationship determining unit 302 (step S302). In the process shown in step S302, the evaluation processing unit 303 obtains a state when the likelihood of occurrence in the relationship is maximum (or substantially maximum), for example. In this case, the evaluation processing unit 303 estimates a state most likely to occur in the relationship.
- the relationship determination unit 302 can be realized using the function of the relationship determination unit 109 in FIG. 1 or the function of the relationship determination unit 210 in FIG.
- the evaluation processing unit 303 can be realized by using the function of the evaluation processing unit 103 in FIG. 1 or the function of the evaluation processing unit 203 in FIG. Therefore, the information processing apparatus 301 can be realized using the function of the information processing apparatus 101 in FIG. 1 or the function of the information processing apparatus 201 in FIG.
- the information processing apparatus 301 According to the information processing apparatus 301 according to the third embodiment, it is possible to provide information that is a basis for estimating a state related to a target with high accuracy. This reason is the same as the reason described in the first embodiment.
- FIG. 8 is a block diagram showing the configuration of the information processing apparatus 401 according to the fourth embodiment of the present invention.
- the information processing apparatus 401 includes a relationship determination unit 402 and an evaluation processing unit 403.
- the information processing apparatus 401 is communicably connected to an estimation apparatus (simulation apparatus) 404.
- the estimation device 404 estimates a plurality of scenarios representing a mode in which a state (state information) related to an object changes.
- the estimation device 404 estimates a plurality of the scenarios according to the assimilation processing described with reference to steps S101 to S105 (FIG. 2) (or estimates the scenarios probabilistically).
- the estimation device 404 can output to the information processing device 401 the likelihood (likelihood) that the scenario occurs with respect to the observation information observed for the object and the scenario.
- the information processing apparatus 401 can be realized using, for example, the function of the simulation unit 102 in FIG. 1 or the function of the simulation unit 202 in FIG.
- FIG. 9 is a flowchart showing the flow of processing in the information processing apparatus 401 according to the fourth embodiment.
- the information processing apparatus 401 inputs a set configured by combining the likelihood of the scenario estimated by the estimation apparatus 404 with respect to the target and the state information indicating the state regarding the target in the scenario. In this case, the information processing apparatus 401 inputs a plurality of sets related to a plurality of scenarios estimated by the estimation apparatus 404.
- the relationship determination unit 402 sets a first condition that at least a part of the plurality of input sets is not distant from other sets or is similar to the other sets in the distribution of at least some sets.
- a satisfying set is selected (step S401).
- the distance represents a mathematical distance.
- the relationship determination unit 402 when calculating the center for all the input sets, takes an average position as described with reference to FIG. 3 (the attention described with reference to FIG. 3). (Example of position) is calculated.
- the relationship determining unit 402 may calculate the center by executing a process similar to the process executed by the relationship determining unit 109, for example. In this case, the relationship determination unit 402 selects a set including state information satisfying a predetermined constraint condition from the plurality of input sets, and calculates a center for the selected set. In this case, the calculated center represents a mathematical (or mathematical) center.
- the relationship determination unit 402 selects a set close to the calculated center. For example, the relationship determining unit 402 determines whether or not the center is separated by “3 ⁇ ⁇ ” ( ⁇ represents a standard deviation) or more, and selects a set within “3 ⁇ ⁇ ” from the center. Therefore, the relationship determination unit 402 selects a set that satisfies the first condition that the distance is not far from another set or is similar to the other set.
- the relationship determination unit 402 creates a cluster according to clustering as a process for suppressing the influence of the outlier value, and selects a cluster having a large number of elements belonging to the created cluster, thereby obtaining the first condition. You may select a set that satisfies Therefore, the process of selecting a set that satisfies the first condition is not limited to the above-described example.
- the evaluation processing unit 403 obtains state information corresponding to the likelihood of occurrence in a set that satisfies a predetermined selection condition among the sets selected by the relationship determining unit 402 (step S402).
- the predetermined selection condition is a condition that the likelihood of occurrence is a large value.
- the relationship determining unit 402 obtains state information when the likelihood is high in a set where the likelihood is high.
- the evaluation processing unit 403 obtains a state that is likely to occur in the set selected by the relationship determination unit 402.
- the evaluation processing unit 403 displays state information corresponding to the likelihood of occurrence in the set that is most likely to occur among the sets selected by the relationship determination unit 402. You may ask for it. In this case, the evaluation processing unit 403 obtains the most likely state in the set selected by the relationship determining unit 402.
- the relationship determining unit 402 can be realized by using the function of the relationship determining unit 109 in FIG. 1 or the function of the relationship determining unit 210 in FIG.
- the evaluation processing unit 403 can be realized by using the function of the evaluation processing unit 103 in FIG. 1 or the function of the evaluation processing unit 203 in FIG. Therefore, the information processing apparatus 401 can be realized using the function of the information processing apparatus 101 in FIG. 1 or the function of the information processing apparatus 201 in FIG.
- the information processing apparatus 401 it is possible to provide information that is a basis for estimating a state related to a target with high accuracy.
- the reason for this is that a set that is likely to occur among a plurality of sets configured by combining the likelihood of occurrence of a plurality of scenarios and state information representing the state of the object in the scenario is set to the center of the plurality of sets. It is because it selects based on. Since the information processing apparatus 401 can remove a set that is out of the distribution related to the plurality of sets from the plurality of sets, it is possible to provide information that is a basis for estimating the state related to the object with high accuracy.
- FIG. 10 is a block diagram schematically showing a hardware configuration example of a calculation processing apparatus capable of realizing the information processing apparatus according to each embodiment of the present invention.
- the computing device 20 includes a central processing unit (Central_Processing_Unit, hereinafter referred to as “CPU”) 21, a memory 22, a disk 23, a nonvolatile recording medium 24, and a communication interface (hereinafter referred to as “communication IF”) 27.
- CPU central processing unit
- the calculation processing device 20 may be connectable to the input device 25 and the output device 26.
- the calculation processing device 20 can transmit / receive information to / from other calculation processing devices and communication devices via the communication IF 27.
- the non-volatile recording medium 24 is a computer-readable, for example, compact disc (Compact_Disc) or digital versatile disc (Digital_Versatile_Disc).
- the nonvolatile recording medium 24 may be a universal serial bus memory (USB memory), a solid state drive (Solid_State_Drive), or the like.
- the non-volatile recording medium 24 retains such a program without being supplied with power, and can be carried.
- the nonvolatile recording medium 24 is not limited to the above-described medium. Further, the program may be carried via the communication IF 27 and the communication network instead of the nonvolatile recording medium 24.
- the CPU 21 copies a software program (computer program: hereinafter simply referred to as “program”) stored in the disk 23 to the memory 22 and executes arithmetic processing.
- the CPU 21 reads data necessary for program execution from the memory 22. When the display is necessary, the CPU 21 displays the output result on the output device 26. When inputting a program from the outside, the CPU 21 reads the program from the input device 25.
- the CPU 21 executes an information processing program (FIG. 2, FIG. 5, FIG. 7 or FIG. 7) in the memory 22 corresponding to the function (processing) represented by each unit shown in FIG. 1, FIG. 4, FIG. 6, or FIG. FIG. 9) is interpreted and executed.
- the CPU 21 sequentially executes the processes described in the above embodiments of the present invention.
- the present invention can also be realized by such an information processing program. Furthermore, it can be understood that the present invention can also be realized by a computer-readable non-volatile recording medium in which the information processing program is recorded.
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Physics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Software Systems (AREA)
- Algebra (AREA)
- Probability & Statistics with Applications (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Operations Research (AREA)
- Databases & Information Systems (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Description
対象に関する状態を表す状態情報と、前記状態情報が変化する態様を表すシナリオの生じやすさとの関係性を決定する関係性決定手段と、
前記関係性決定手段が決定した前記関係性に基づいて、前記生じやすさが所定の選択条件を満たす場合の前記状態情報を求める評価処理手段と
を備える。
計算処理装置によって、対象に関する状態を表す状態情報と、前記状態情報が変化する態様を表すシナリオの生じやすさとの関係性を決定し、決定した前記関係性に基づいて、前記生じやすさが所定の選択条件を満たす場合の前記状態情報を求める。
対象に関する状態を表す状態情報と、前記状態情報が変化する態様を表すシナリオの生じやすさとの関係性を決定する関係性決定機能と、
前記関係性決定機能において決定された前記関係性に基づいて、前記生じやすさが所定の選択条件を満たす場合の前記状態情報を求める評価処理機能と
をコンピュータに実現させる。
図1を参照しながら、本発明の第1の実施形態に係る情報処理装置101が有する構成について詳細に説明する。図1は、本発明の第1の実施形態に係る情報処理装置101が有する構成を示すブロック図である。
次に、上述した第1の実施形態を基本とする本発明の第2の実施形態について説明する。
次に、本発明の第3の実施形態について説明する。
次に、本発明の第4の実施形態について説明する。
上述した本発明の各実施形態に係る情報処理装置を、1つの計算処理装置(コンピュータ)を用いて実現するハードウェア資源の構成例について説明する。但し、係る情報処理装置は、物理的または機能的に少なくとも2つの計算処理装置を用いて実現されてもよい。また、係る情報処理装置は、専用の装置として実現されてもよい。
102 シミュレーション部
103 評価処理部
104 データ取得部
105 同化計算部
106 予測部
107 計算終了判定部
108 評価値計算部
109 関係性決定部
110 状態算出部
151 観測装置
152 入力装置
153 出力装置
154 観測情報記憶部
155 設定情報記憶部
201 情報処理装置
202 シミュレーション部
203 評価処理部
204 データ取得部
205 同化計算部
206 予測部
207 計算終了判定部
208 統計処理部
209 評価値計算部
210 関係性決定部
211 状態算出部
212 乱数種情報
301 情報処理装置
302 関係性決定部
303 評価処理部
304 推定装置
401 情報処理装置
402 関係性決定部
403 評価処理部
404 推定装置
20 計算処理装置
21 CPU
22 メモリ
23 ディスク
24 不揮発性記録媒体
25 入力装置
26 出力装置
27 通信IF
501 観測情報
502 観測モデル
503 システムモデル
504 状態推定
Claims (10)
- 対象に関する状態を表す状態情報と、前記状態情報が変化する態様を表すシナリオの生じやすさとの関係性を決定する関係性決定手段と、
前記関係性決定手段が決定した前記関係性に基づいて、前記生じやすさが所定の選択条件を満たす場合の前記状態情報を求める評価処理手段と
を備える情報処理装置。 - 前記評価処理手段は、前記生じやすさが最大、または、略最大である場合の前記状態情報を求める
請求項1に記載の情報処理装置。 - 前記生じやすさに基づき、前記シナリオに含まれている期間における前記生じやすさを算出する計算手段
をさらに備え、
前記シナリオは、複数のタイミングにおける前記生じやすさを含む
請求項1または請求項2に記載の情報処理装置。 - 前記関係性決定手段は、前記シナリオに含まれる前記状態情報と、前記シナリオの生じやすさとのセットのうち、前記状態が所定の条件を満たしているセットを用いて前記関係性を決定する
請求項1乃至請求項3のいずれかに記載の情報処理装置。 - 前記関係性決定手段は、前記シナリオに含まれる前記状態情報と、前記シナリオの生じやすさとのセットを複数のクラスタに分類し、分類したクラスタのうち該クラスタに属している前記セットの個数が多いクラスタを用いて前記関係性を決定する
請求項1乃至請求項3のいずれかに記載の情報処理装置。 - 前記関係性は、上に凸な関数、または、ガウス関数である
請求項1乃至請求項5のいずれかに記載の情報処理装置。 - 処理手段
をさらに備え、
前記計算手段は、前記期間に関する生じやすさを、前記複数のシナリオに関する平均的な生じやすさを算出し、
前記処理手段は、前記期間に関する生じやすさを重みとした前記状態情報の第1加重平均と、前記期間に関する生じやすさを重みとした前記平均的な生じやすさの第2加重平均とを算出し、
前記関係性決定手段は、前記処理手段が算出した前記第1加重平均と、前記第2加重平均とに基づき、前記関係性を決定する
請求項3に記載の情報処理装置。 - 処理手段
をさらに備え、
前記計算手段は、前記期間に関する生じやすさを、前記複数のシナリオに関する平均的な生じやすさを算出し、
前記処理手段は、前記期間に関する生じやすさが大きい場合の、前記状態情報と、前記平均的な生じやすさとを選択し、
前記関係性決定手段は、前記処理手段が選択した前記状態情報と、前記処理手段が選択した前記平均的な生じやすさとに基づき、前記関係性を決定する
請求項3に記載の情報処理装置。 - 計算処理装置によって、対象に関する状態を表す状態情報と、前記状態情報が変化する態様を表すシナリオの生じやすさとの関係性を決定し、決定した前記関係性に基づいて、前記生じやすさが所定の選択条件を満たす場合の前記状態情報を求める情報処理方法。
- 対象に関する状態を表す状態情報と、前記状態情報が変化する態様を表すシナリオの生じやすさとの関係性を決定する関係性決定機能と、
前記関係性決定機能において決定された前記関係性に基づいて、前記生じやすさが所定の選択条件を満たす場合の前記状態情報を求める評価処理機能と
をコンピュータに実現させる情報処理プログラムが記録された記録媒体。
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2018564511A JP6763444B2 (ja) | 2017-01-24 | 2018-01-17 | 情報処理装置、情報処理方法、及び、情報処理プログラム |
| US16/477,251 US20190362258A1 (en) | 2017-01-24 | 2018-01-17 | Information processing apparatus, information processing method, and non-transitory recording medium |
| EP18745052.3A EP3575992A4 (en) | 2017-01-24 | 2018-01-17 | INFORMATION PROCESSING DEVICE, INFORMATION PROCESSING METHOD, AND RECORDING MEDIUM ON WHICH AN INFORMATION PROCESSING PROGRAM IS RECORDED |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2017010441 | 2017-01-24 | ||
| JP2017-010441 | 2017-01-24 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2018139300A1 true WO2018139300A1 (ja) | 2018-08-02 |
Family
ID=62978265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2018/001160 Ceased WO2018139300A1 (ja) | 2017-01-24 | 2018-01-17 | 情報処理装置、情報処理方法、及び、情報処理プログラムが記録された記録媒体 |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US20190362258A1 (ja) |
| EP (1) | EP3575992A4 (ja) |
| JP (1) | JP6763444B2 (ja) |
| WO (1) | WO2018139300A1 (ja) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| IL289666A (en) * | 2022-01-06 | 2023-08-01 | SADEH Yuval | Assesses variety for crop yield |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS5340228B2 (ja) | 1974-09-02 | 1978-10-26 | ||
| JP2013200683A (ja) * | 2012-03-23 | 2013-10-03 | Nippon Telegr & Teleph Corp <Ntt> | 状態追跡装置、方法、及びプログラム |
| WO2014141344A1 (ja) * | 2013-03-14 | 2014-09-18 | 日本電気株式会社 | データ予測装置 |
| JP2016004525A (ja) * | 2014-06-19 | 2016-01-12 | 株式会社日立製作所 | データ分析システム及びデータ分析方法 |
| WO2016031174A1 (ja) | 2014-08-27 | 2016-03-03 | 日本電気株式会社 | シミュレーション装置、シミュレーション方法、および、記憶媒体 |
| JP2016071383A (ja) * | 2014-09-26 | 2016-05-09 | インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Machines Corporation | 情報処理装置、プログラム、及び、情報処理方法 |
| JP2017010441A (ja) | 2015-06-25 | 2017-01-12 | ローム株式会社 | 半導体集積回路の回路シミュレーション方法及び半導体集積回路 |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6829628B2 (en) * | 2001-05-02 | 2004-12-07 | Portalplayer, Inc. | Random number generation method and system |
| US9818136B1 (en) * | 2003-02-05 | 2017-11-14 | Steven M. Hoffberg | System and method for determining contingent relevance |
| WO2008143026A1 (ja) * | 2007-05-24 | 2008-11-27 | Nec Corporation | スループット推定方法及びシステム |
| JP6120665B2 (ja) * | 2013-04-26 | 2017-04-26 | オリンパス株式会社 | 撮像装置、画像処理装置、画像処理方法及び画像処理プログラム |
| US20180020622A1 (en) * | 2016-07-25 | 2018-01-25 | CiBo Technologies Inc. | Agronomic Database and Data Model |
-
2018
- 2018-01-17 JP JP2018564511A patent/JP6763444B2/ja not_active Expired - Fee Related
- 2018-01-17 US US16/477,251 patent/US20190362258A1/en not_active Abandoned
- 2018-01-17 EP EP18745052.3A patent/EP3575992A4/en active Pending
- 2018-01-17 WO PCT/JP2018/001160 patent/WO2018139300A1/ja not_active Ceased
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS5340228B2 (ja) | 1974-09-02 | 1978-10-26 | ||
| JP2013200683A (ja) * | 2012-03-23 | 2013-10-03 | Nippon Telegr & Teleph Corp <Ntt> | 状態追跡装置、方法、及びプログラム |
| WO2014141344A1 (ja) * | 2013-03-14 | 2014-09-18 | 日本電気株式会社 | データ予測装置 |
| JP2016004525A (ja) * | 2014-06-19 | 2016-01-12 | 株式会社日立製作所 | データ分析システム及びデータ分析方法 |
| WO2016031174A1 (ja) | 2014-08-27 | 2016-03-03 | 日本電気株式会社 | シミュレーション装置、シミュレーション方法、および、記憶媒体 |
| JP2016071383A (ja) * | 2014-09-26 | 2016-05-09 | インターナショナル・ビジネス・マシーンズ・コーポレーションInternational Business Machines Corporation | 情報処理装置、プログラム、及び、情報処理方法 |
| JP2017010441A (ja) | 2015-06-25 | 2017-01-12 | ローム株式会社 | 半導体集積回路の回路シミュレーション方法及び半導体集積回路 |
Non-Patent Citations (1)
| Title |
|---|
| See also references of EP3575992A4 |
Also Published As
| Publication number | Publication date |
|---|---|
| JP6763444B2 (ja) | 2020-09-30 |
| EP3575992A1 (en) | 2019-12-04 |
| EP3575992A4 (en) | 2020-02-19 |
| US20190362258A1 (en) | 2019-11-28 |
| JPWO2018139300A1 (ja) | 2019-11-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Van Eupen et al. | The impact of data quality filtering of opportunistic citizen science data on species distribution model performance | |
| US20210150717A1 (en) | Analyzing operational data influencing crop yield and recommending operational changes | |
| Van Beveren et al. | How catch underreporting can bias stock assessment of and advice for northwest Atlantic mackerel and a possible resolution using censored catch | |
| Tabas et al. | Variational Bayesian dropout with a Gaussian prior for recurrent neural networks application in rainfall–runoff modeling | |
| Orhan | Particle filtering | |
| US20200073915A1 (en) | Information processing apparatus, information processing system, and information processing method | |
| KR102120150B1 (ko) | 뉴럴 네트워크를 이용한 변분 추론 학습 방법 및 학습 장치, 그리고 이를 이용한 변분 추론 방법 및 장치 | |
| WO2024059300A1 (en) | Uncertainty prediction models | |
| CN119151462B (zh) | 一种耕地土壤数据自动化审核分析方法 | |
| US20230214668A1 (en) | Hyperparameter adjustment device, non-transitory recording medium in which hyperparameter adjustment program is recorded, and hyperparameter adjustment program | |
| CN117892638A (zh) | 一种利用条件概率函数的干旱形成时间预测方法及系统 | |
| Bessell-Browne et al. | Management strategy evaluation of static and dynamic harvest control rules under long-term changes in stock productivity: A case study from the SESSF | |
| Morandage et al. | Bayesian inference of root architectural model parameters from synthetic field data | |
| CN117121028A (zh) | 用于概率时空预测的方法、系统及计算机可读介质 | |
| WO2018139300A1 (ja) | 情報処理装置、情報処理方法、及び、情報処理プログラムが記録された記録媒体 | |
| CN112597827A (zh) | 一种基于大数据的植物物候期预测方法和系统 | |
| CN116579521A (zh) | 产量预测时间窗口确定方法、装置、设备及可读存储介质 | |
| JP6908111B2 (ja) | 予測装置、予測方法、予測プログラム、及び、遺伝子推定装置 | |
| WO2018139301A1 (ja) | 情報処理装置、情報処理方法、及び、情報処理プログラムが記録された記録媒体 | |
| Dahia et al. | A Systematic Review of Cancer Burden Forecasting Models: Evaluating Efficacy for Long-Term Predictions Using Annual Data | |
| Dawod et al. | Forecasting Short-Term Wind Speed using ARIMA and Convolution Neural Network Models at International Airports in Saudi Arabia | |
| RU2622858C1 (ru) | Способ оценки информации об эффективности функционирования системы и устройство на его основе для решения задач управления, контроля и диагностики | |
| Dekker et al. | Estimating Non-Stationary Extreme-Value Probability Distribution Shifts and Their Parameters Under Climate Change Using L-Moments and L-Moment Ratio Diagrams: A Case Study of Hydrologic Drought in the Goat River Near Creston, British Columbia. Hydrology 2024, 11, 154 | |
| JP6787561B2 (ja) | 情報処理装置、方法、及びプログラム | |
| JP7385931B2 (ja) | 情報処理装置、情報処理方法、およびプログラム |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 18745052 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 2018564511 Country of ref document: JP Kind code of ref document: A |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| ENP | Entry into the national phase |
Ref document number: 2018745052 Country of ref document: EP Effective date: 20190826 |
|
| WWW | Wipo information: withdrawn in national office |
Ref document number: 2018745052 Country of ref document: EP |