Information processing device, information processing method, information processing system, and program
The information processing device addresses the computational cost issue in prediction technologies by updating model reliability based on loss values and initializing at different timings, ensuring efficient and accurate predictions in dynamic environments.
Patent Information
- Application Number
- PCT/JP2023/017066
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-01
- Publication Date
- 2024-11-07
AI Technical Summary
Existing prediction technologies that refer to multiple experts' values face increased computational costs, particularly in memory capacity, while adapting to changing environments.
An information processing device and method that acquires output values from multiple models, derives an optimal solution based on model reliability, and updates reliability using loss values, with the reliability of models initialized at different timings to manage calculation costs.
This approach effectively suppresses calculation costs while maintaining decision-making accuracy by dynamically resetting model reliability, allowing for efficient prediction in changing environments.
Smart Images

Figure JP2023017066_07112024_PF_FP_ABST
Abstract
Description
Information processing device, information processing method, information processing system, and program
[0001] The present invention relates to an information processing device, an information processing method, an information processing system, and a program.
[0002] A sequential decision-making technique is known in which a process of deriving a prediction (decision-making result) regarding demand or supply, etc., observing the execution results of the derived prediction, and deriving a further prediction based on the observed results is sequentially repeated.
[0003] For example, Patent Document 1 describes a technology for predicting traffic demand within a building, which includes a plurality of experts who refer to different data contained in a feature database to predict categories of traffic demand and generate predicted values, and one of the predicted values is adopted as the prediction result.
[0004] JP 2014-172718 A
[0005] Prediction techniques that refer to predicted values from multiple experts, such as those described in Patent Document 1, have the advantage of being able to make predictions while appropriately adapting to changing environments, but they have had the problem of increasing calculation costs, such as the amount of calculation, calculation time, and memory capacity.
[0006] One aspect of the present invention has been made in consideration of the above-mentioned problems, and one of its objectives is to provide an optimization technique that references output values from multiple experts (models) while suppressing computational costs.
[0007] An information processing device according to one aspect of the present invention includes an acquisition means for acquiring output values obtained from each of a plurality of models and loss values corresponding to each output value, a derivation means for deriving an optimal solution based on each output value acquired by the acquisition means and the reliability of each model, and an update means for updating the reliability of each model by referring to each loss value acquired by the acquisition means, wherein the update means initializes the reliability of each of a plurality of models constituting at least a portion of the plurality of models at mutually different times.
[0008] An information processing method according to one aspect of the present invention includes an information processing device acquiring output values obtained from each of a plurality of models and loss values corresponding to each output value, deriving an optimal solution based on the acquired output values and the reliability of each model, and updating the reliability for each model by referring to the acquired loss values, wherein in the updating step, the reliability of each of a plurality of models constituting at least a portion of the plurality of models is initialized at mutually different times.
[0009] An information processing program according to one aspect of the present invention is a program that causes a computer to function as an information processing device, and the program causes the computer to acquire output values obtained from each of a plurality of models and loss values corresponding to each output value, derive an optimal solution based on the acquired output values and the reliability of each model, and update the reliability of each model by referring to the acquired loss values, and in the updating step, initialize the reliability of each of a plurality of models that constitute at least a portion of the plurality of models at mutually different times.
[0010] An information processing system according to one aspect of the present invention is an information processing system including an information processing device and a terminal device, wherein the information processing device comprises: an acquisition means for acquiring output values obtained from each of a plurality of models and loss values corresponding to each output value; a derivation means for deriving an optimal solution based on each output value acquired by the acquisition means and the reliability of each model; and an update means for updating the reliability of each model by referring to each loss value acquired by the acquisition means, wherein the update means initializes the reliability of each of a plurality of models constituting at least a portion of the plurality of models at mutually different times; and the terminal device comprises an execution means for executing the optimal solution derived by the information processing device; and a loss value acquisition means for acquiring the loss value obtained by executing the optimal solution.
[0011] According to one aspect of the present invention, it is possible to provide an optimization technique that references output values from multiple experts while suppressing calculation costs.
[0012] FIG. 1 is a block diagram showing the configuration of an information processing device according to exemplary embodiment 1. FIG. 2 is a flow diagram showing the flow of an information processing method according to exemplary embodiment 1. FIG. 3 is a diagram for explaining processing by the information processing device according to exemplary embodiment 1. FIG. 4 is a block diagram showing the configuration of an information processing system according to exemplary embodiment 1. FIG. 5 is a flow diagram showing the flow of processing by the information processing system according to exemplary embodiment 1. FIG. 6 is a block diagram showing the configuration of an information processing device according to exemplary embodiment 2. FIG. 7 is a diagram for explaining processing by the information processing device according to an application example of exemplary embodiment 2. FIG. 8 is a diagram showing an example of information referenced by an information processing device according to an application example of exemplary embodiment 2. FIG. 9 is a block diagram showing the configuration of a computer functioning as an information processing device according to each exemplary embodiment.
[0013] [First Exemplary Embodiment] A first exemplary embodiment of the present invention will be described in detail with reference to the drawings. This exemplary embodiment is a basic form of the exemplary embodiments described below.
[0014] <Overview of Information Processing Device 1> The information processing device 1 according to this exemplary embodiment is an information processing device that acquires predicted values from multiple experts and makes decisions using the acquired predicted values. Furthermore, the information processing device 1 according to this exemplary embodiment sequentially acquires predicted values and makes decisions. For example, in round t, a predicted value P1t is acquired from expert 1 and a predicted value P2t is acquired from expert 2, and a decision-making result (t) for round t is derived by referring to the acquired predicted values P1t and P2t. The information processing device 1 then updates parameters for decision-making, acquires predicted values for the next round, and derives a decision-making result for the next round. Note that t is an index representing the number of repetitions and can also be interpreted as an index indicating timing.
[0015] In this exemplary embodiment, the "expert" may specifically be hardware, such as a predicted value derivation device that outputs predicted values, software, such as a predicted value derivation algorithm that outputs predicted values, or a living organism (e.g., a human) that outputs predicted values using some method. Furthermore, the information processing device 1 according to this exemplary embodiment may be configured to include an "expert," or may be configured to acquire predicted values from an external "expert." The "expert" may also be referred to as a "model" or "agent."
[0016] Furthermore, in this exemplary embodiment, the "predicted value" may be, for example, a predicted value related to supply or demand, or a predicted value related to other cases (events). For example, the "predicted value" in this exemplary embodiment may be a predicted value related to some parameter referenced by the information processing device 1. The information processing device 1 according to this exemplary embodiment can be applied to the overall process of making decisions regarding a target event by referring to predicted values from multiple experts. Note that, although this exemplary embodiment describes a case where the output value of an expert is a "predicted value," the output value of an expert is not necessarily limited to a predicted value. The output value of an expert may be any value that can define a loss, as described below. For example, the output value of an expert may be the output of a generative model or a control model.
[0017] Furthermore, in this exemplary embodiment, "intention" refers to some information related to a target event, and is not limited to the intention of a living organism (person). For example, in an application scenario in which demand for a target product is predicted, a predicted value of future demand is an example of an "intention" determined by the information processing device 1 according to this exemplary embodiment, or an example of a "decision-making result" derived by the information processing device 1. The information processing device 1 according to this exemplary embodiment can also be expressed as a decision-making device or a decision-making result derivation device. The "decision-making result" is also called an "optimization solution" or an "optimization result."
[0018] In this exemplary embodiment, a loss value is provided corresponding to a predicted value provided by each of the multiple experts. For example, a loss value L1 is provided corresponding to a predicted value P1 provided by expert 1, and a loss value L2 is provided corresponding to a predicted value P2 provided by expert 2. Each predicted value provided by the experts and the corresponding loss value are acquired by the information processing device 1 and are referred to in decision-making processing or other information processing.
[0019] In this exemplary embodiment, the loss value can be expressed as the difference between a predicted value and an observed value (actually measured value), for example, but this does not limit the exemplary embodiment. The loss value may also be the difference between a predicted value and another predetermined value. The loss value may also be an estimated value related to the loss. The term "loss value" may also include the concept of "reward." For example, the loss value may be expressed as the reward value with the sign reversed (the reward value multiplied by a negative constant). Therefore, the loss value according to this exemplary embodiment may be read as the reward value. The information referenced to calculate the loss value and the specific algorithm for calculating the loss value do not limit the exemplary embodiment.
[0020] <Configuration of information processing device 1> Next, the configuration of the information processing device 1 will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the information processing device 1. As shown in Fig. 1, the information processing device 1 includes an acquisition unit 11 and a derivation unit 12.
[0021] (Acquisition Unit 11) The acquisition unit 11 acquires predicted values obtained from each of the multiple experts and loss values corresponding to each predicted value. Here, as described above, the "predicted value" may be, for example, a predicted value related to supply and demand, or a predicted value related to other cases (events). Furthermore, the "predicted value" may be a predicted value related to some parameter referenced by the information processing device 1.
[0022] Furthermore, the "loss value corresponding to each predicted value" refers to a loss value that has some kind of information processing relevance to each predicted value. For example, the loss value L1 may indicate: - the difference between the predicted value P1 and the observed value; or - the difference between the decision-making result derived by the decision-making process that references the predicted value P1 and the observed value. Furthermore, as described above, the concept of a loss value may include the concept of a reward value.
[0023] The acquisition unit 11 may acquire the predicted value and the loss value from the same acquisition source or from different acquisition sources. Furthermore, the acquisition unit 11 may be configured to acquire the predicted value and the loss value for each round in the sequential decision-making process, but this is not a limitation of the present exemplary embodiment.
[0024] (Derivation unit 12) The derivation unit 12 performs the following processes: deriving a decision-making result according to each predicted value acquired by the acquisition unit 11 and the reliability of each expert; and updating the reliability of each expert by referring to each loss value acquired by the acquisition unit 11.
[0025] Here, the reliability is an index indicating the degree to which the prediction value by each expert is reflected in the decision-making process. As an example, the reliability can be expressed as a relative weight calculated for the prediction value by each expert, but this is not intended to limit the present exemplary embodiment.
[0026] The derivation unit 12 also resets the reliability of each of the multiple experts constituting at least a portion of the multiple experts at different times. For example, in a situation where the acquisition unit 11 acquires a predicted value P1 of Expert 1, a predicted value P2 of Expert 2, and a predicted value P3 of Expert 3, the derivation unit 12 resets the reliability w1 of Expert 1 and the reliability w2 of Expert 2 at different times (rounds). Here, "resetting" refers to, for example, a process of setting the reliability of an object to 0, but this does not limit the present exemplary embodiment. "Resetting" may refer to any process of setting the reliability of an object to a predetermined value. Note that "resetting" is also referred to as "initialization." The derivation unit 12 according to this embodiment is an example of the "deriving means" and "updating means" in the claims.
[0027] As described above, the information processing device 1 according to this exemplary embodiment derives a decision-making result (optimal solution) according to the predicted values (output values) provided by each of the multiple experts (models) and the reliability of each expert (model), and resets the reliability of each of the multiple experts (models) that constitute at least a portion of the multiple experts (models) at different times (rounds). In this way, the information processing device 1 according to this exemplary embodiment resets the reliability of each of the multiple experts (models) that constitute at least a portion of the multiple experts (models) at different times, thereby making it possible to suppress an increase in calculation costs while suppressing a decrease in the accuracy of the decision-making result (optimal solution).
[0028] <Flow of Information Processing Method S1> Next, the flow of the information processing method S1 according to the present exemplary embodiment 1 will be described with reference to Fig. 2. Fig. 2 is a flow chart showing the flow of the information processing method S1.
[0029] (Step S11) In step S11, the obtaining unit 11 obtains the predicted values obtained from each of the multiple experts and the loss values corresponding to each predicted value.
[0030] (Step S12) In step S12, the derivation unit 12 performs the following processes: deriving a decision-making result according to each predicted value acquired by the acquisition unit 11 and the reliability of each expert; and updating the reliability of each expert by referring to each loss value acquired by the acquisition unit 11.
[0031] In step S12, the derivation unit 12 resets the reliability of each of the experts constituting at least a portion of the plurality of experts at different times.
[0032] In other words, the derivation unit 12 resets the reliability of a certain expert in step S12 in a certain round, and resets the reliability of another expert in step S12 in another round.
[0033] The information processing device 1 performs the processes of steps S11 and S12 in a certain round, and then performs the processes of steps S11 and S12 in the next round.
[0034] 3 is a diagram for schematically illustrating sequential decision-making processing by the information processing method S1 according to this exemplary embodiment. As shown in FIG. 3 , in a certain round, the information processing device 1 acquires, for multiple experts, predicted values and loss values associated with each expert. Then, the information processing device 1 derives a decision-making result by referring to the acquired predicted values and loss values. The derived decision-making result is then executed, thereby obtaining a loss value corresponding to the next round. The loss value and the predicted value corresponding to the loss value are provided to the information processing device 1 and are referenced in the decision-making processing in the next round. In this manner, the information processing device 1 sequentially derives decision-making results.
[0035] As described above, in the information processing method S1 according to this exemplary embodiment, a decision-making result (optimal solution) is derived based on the predicted values (output values) provided by each of a plurality of experts (models) and the reliability of each expert (model), and the reliability of each of the plurality of experts (models) constituting at least a portion of the plurality of experts (models) is reset at mutually different times (rounds). In this way, the information processing device 1 according to this exemplary embodiment resets the reliability of each of the plurality of experts (models) constituting at least a portion of the plurality of experts (models) at mutually different times, thereby suppressing a decrease in the accuracy of the decision-making result (optimal solution) while suppressing an increase in calculation costs.
[0036] <Configuration of Information Processing System 100> Next, the configuration of the information processing system 100 according to this exemplary embodiment will be described with reference to Fig. 4. Fig. 4 is a block diagram showing the configuration of the information processing system 100. As shown in Fig. 4, the information processing device 100 includes an information processing device 1 and a terminal device 2 that are communicably connected to each other. Since the components of the information processing device 1 have been described above, description thereof will be omitted here.
[0037] 4, the terminal device 2 includes an execution unit 21 and a loss value acquisition unit 22. The execution unit 21 executes the decision-making result derived by the information processing device 1 or a process corresponding to the decision-making result. As an example, if the decision-making result predicts X units of product A as today's demand, the execution unit 21 places an order for X units of product A.
[0038] The loss value acquiring unit 22 acquires a loss value corresponding to each of the prediction values of the multiple experts referred to by the information processing device 1. As an example, the loss value acquiring unit 22 acquires: a prediction value P1 provided by expert 1, which is the difference between the prediction value P1 of today's demand for product A and the number of units of product A sold today, as loss L1; and a prediction value P2 provided by expert 2, which is the difference between the prediction value P2 of today's demand for product A and the number of units of product A sold today, as loss L1.
[0039] The acquired loss value is referred to in the information processing device 1 in the process of updating the reliability of each expert in the next round (tomorrow in this example).
[0040] <Flow of Information Processing Method S100> Next, the flow of the information processing method S100 according to the present exemplary embodiment 1 will be described with reference to Fig. 5. Fig. 5 is a flow diagram showing the flow of the information processing method S100 executed by the information processing system 100.
[0041] (Steps S11-1, S12-1) As shown in FIG. 4, in step S11-1, the acquisition unit 11 acquires predicted values for the current round (round t=1) obtained from each of the multiple experts, and loss values corresponding to each predicted value.
[0042] In step S12-1, the derivation unit 12 performs the following processes: deriving a decision-making result in accordance with each predicted value acquired by the acquisition unit 11 in step S11-1 and the reliability of each expert; and updating the reliability of each expert by referring to each loss value acquired by the acquisition unit 11 in step S12-1. The decision-making result derived in step S12-1 is provided to the terminal device 2.
[0043] In step S12-1, the derivation unit 12 resets the reliability of each of the experts constituting at least a portion of the plurality of experts at different times.
[0044] In other words, in step S12-1 in the current round, the derivation unit 12 resets the reliability of a certain expert (for example, expert 1), but does not reset the reliability of other experts (for example, experts 2 and 3).
[0045] (Steps S21-1, S22-1) In step S21-1, the execution unit 21 of the terminal device 2 executes the decision-making result determined in step S12-1 or a process corresponding to the decision-making result. In step S22-1, the loss value acquisition unit 22 of the terminal device 2 acquires the loss value obtained as a result of the execution and provides it to the information processing device 1.
[0046] (Steps S11-2 and S12-2) In step S11-2, the acquiring unit 11 acquires predicted values for the current round (round t=2) obtained from each of the experts and loss values corresponding to each predicted value. Here, each predicted value is, for example, the loss value acquired by the loss value acquiring unit 22 in step S22-1.
[0047] In step S12-2, the derivation unit 12 performs the following processes: deriving a decision-making result in accordance with each predicted value acquired by the acquisition unit 11 in step S11-2 and the reliability of each expert, and updating the reliability of each expert by referring to each loss value acquired by the acquisition unit 11 in step S11-2. The decision-making result derived in step S12-2 is provided to the terminal device 2 and executed in step S21-2.
[0048] In step S12-2, the derivation unit 12 resets the reliability of each of the experts constituting at least a portion of the plurality of experts at different times.
[0049] In other words, in step S12-2 in the current round, the derivation unit 12 resets the reliability of a certain expert (for example, expert 2) but does not reset the reliability of other experts (for example, experts 1 and 3).
[0050] As described above, in the information processing method S100 according to this exemplary embodiment, a decision-making result (optimal solution) is derived based on the predicted values (output values) provided by each of a plurality of experts (models) and the reliability of each expert (model), and the reliability of each of the plurality of experts (models) constituting at least a portion of the plurality of experts (models) is reset at mutually different times (rounds). In this way, the information processing system 100 according to this exemplary embodiment resets the reliability of each of the plurality of experts (models) constituting at least a portion of the plurality of experts (models) at mutually different times, thereby suppressing a decrease in the accuracy of the decision-making result (optimal solution) while suppressing an increase in calculation costs.
[0051]
[0033] A second exemplary embodiment of the present invention will be described in detail with reference to the drawings. Note that components having the same functions as those described in the first exemplary embodiment are denoted by the same reference numerals, and their description will be omitted as appropriate.
[0052] <Configuration of Information Processing System 100A> The configuration of the information processing system 100A according to this exemplary embodiment will be described with reference to Fig. 6. Fig. 6 is a block diagram showing the configuration of the information processing system 100A. As shown in Fig. 6, the information processing system 100A includes an information processing device 1A and a terminal device 2A. Also, as shown in Fig. 6, the information processing device 1A and the terminal device 2A are configured to be able to communicate with each other via a network N. Here, the specific configuration of the network N does not limit this exemplary embodiment, but as an example, a wireless local area network (LAN), a wired LAN, a wide area network (WAN), a public line network, a mobile data communication network, or a combination of these networks can be used.
[0053] <Configuration of Information Processing Apparatus 1A> The configuration of the information processing apparatus 1A according to this exemplary embodiment will be described with reference to Fig. 6. Fig. 6 is a block diagram showing the configuration of the information processing apparatus 1A.
[0054] As shown in FIG. 6, the information processing device 1A includes a control unit 10A, a storage unit 15A, and a communication unit 16A.
[0055] The communication unit 16A communicates with devices external to the information processing device 1A. As an example, the communication unit 16A communicates with the terminal device 2A. The communication unit 16A transmits data supplied from the control unit 10A to the terminal device 2A, and supplies data received from the terminal device 2A to the control unit 10A.
[0056] (Storage Unit 15A) The storage unit 15A stores various types of information referenced by the control unit 10A and various types of information derived by the control unit 10A. As an example, the storage unit 15A stores: prediction value information PI including prediction values by each of the multiple experts; loss value information LI including loss values corresponding to the prediction values by each of the multiple experts; reliability information CI including the reliability of each of the multiple experts; first decision-making results DR1-1, DR1-2, ... derived by each of multiple first derivation units 121-1, 121-2, ... described below; and a second decision-making result DR2 derived by a second derivation unit 122 described below.
[0057] As described in exemplary embodiment 1, the information processing system 100A according to this exemplary embodiment may be configured to include an “expert” or to obtain predicted values from an “expert” outside the system.
[0058] (Control Unit 10A) As shown in FIG. 6, the control unit 10A includes an acquisition unit 11 and a derivation unit 12.
[0059] (Acquisition Unit 11) The acquisition unit 11 acquires predicted values obtained from each of the multiple experts and loss values corresponding to each predicted value, as in the exemplary embodiment 1. The predicted values and loss values have been described in the exemplary embodiment 1, and therefore a similar description will be omitted. Specific examples of the predicted values and loss values will be described later.
[0060] (Derivation Unit 12) As in the first exemplary embodiment, the derivation unit 12 performs the following operations: deriving a decision-making result according to each predicted value acquired by the acquisition unit 11 and the reliability of each expert; and updating the reliability of each expert by referring to each loss value acquired by the acquisition unit 11.
[0061] Here, the reliability has been explained in the first exemplary embodiment, so the same explanation will be omitted. Specific examples of the reliability will be described later.
[0062] Furthermore, the derivation unit 12 may derive the decision-making result by further referring to a parameter for correcting the reliability. With this configuration, the reliability of each expert can be appropriately corrected using the parameter, and therefore, the information processing device 1 can be operated with a minimum level of accuracy guaranteed, i.e., conservatively.
[0063] Furthermore, similarly to the first exemplary embodiment, the derivation unit 12 resets the reliability of each of the experts constituting at least a portion of the plurality of experts at different times.
[0064] As an example, the derivation unit 12 may calculate the reliability w of the n-th expert among the plurality of experts. n Using a predetermined proportional coefficient d (d is a natural number) and a constant term g (g is an integer), t = d × 2 n It may be configured to reset at timing (round) t expressed as +g (where the symbol x is an operation symbol representing multiplication; the same applies below).
[0065] As another example, the derivation unit 12 may calculate the reliability w of the n-th expert among the plurality of experts. n may be configured to be reset at timing (round) t expressed as t = d × N + n + g, where N is the total number of experts and predetermined constants d (d is a natural number) and g (g is an integer).
[0066] In this way, the information processing system 100A according to this exemplary embodiment resets the reliability of each of the multiple experts constituting at least a portion of the multiple experts at different times, thereby suppressing a decrease in the accuracy of decision-making results while suppressing an increase in computational costs.
[0067] As shown in FIG. 6, the lead-out portion 12 according to this exemplary embodiment includes a plurality of first lead-out portions 12-1, 12-2, . . . and a second lead-out portion 122.
[0068] Here, each first derivation unit 121-j (j is an index for distinguishing the first derivation units from each other) calculates each predicted value (m t) and the reliability of each expert (w1 t ) and the first decision-making result (p t (j)) is derived, and each loss value (l t ) and calculate the confidence level (w1 t ) updated (w1 t+1 (The index for distinguishing each expert from the others is omitted in the parenthesized expression.)
[0069] The second derivation unit 122 also calculates the first decision-making results (p t (j)) and the reliability (w2 t (j)) and derive a second decision-making result according to the above.
[0070] The derivation unit 12 according to this exemplary embodiment performs the above-described hierarchical decision-making process, as an example, using a plurality of first derivation units 12-1, 121-2, ... and a second derivation unit 122, but this is not intended to limit this exemplary embodiment.
[0071] 6, the terminal device 2A includes a control unit 20A, an execution unit 21, and a communication unit 26. As an example, the terminal device 2A can be specifically realized as a checkout terminal located in a store, an inventory management terminal located in a warehouse, or the like, but this does not limit the present exemplary embodiment.
[0072] The communication unit 26 communicates with devices external to the terminal device 2A. As an example, the communication unit 26 communicates with the information processing device 1A. The communication unit 26 transmits data supplied from the control unit 20A to the information processing device 1A, and supplies data received from the information processing device 1A to the control unit 20A.
[0073] The execution unit 21 performs processing on the decision-making result derived by the derivation unit 12 or on the decision-making result. As an example, the execution unit 21 executes the second decision-making result derived by the derivation unit 12. The execution unit 21 may be configured to execute at least some of the multiple first decision-making results derived by the derivation unit 12 instead of or together with the second decision-making result derived by the derivation unit 12.
[0074] 6, the control unit 20A includes a loss value acquisition unit 22 and a loss value providing unit 23. The loss value acquisition unit 22 acquires loss values for each expert obtained as a result of execution by the execution unit 21. The loss value providing unit 23 provides the loss values acquired by the loss value acquisition unit 22 to the information processing device 1A via the communication unit 26.
[0075] 7 is a diagram for schematically illustrating sequential decision-making processing by the information processing system 100A according to this exemplary embodiment. As shown in FIG. 7 , in a certain round, the information processing device 1A acquires predicted values and loss values associated with each expert for multiple experts, and derives a decision-making result by referring to the acquired predicted values and loss values. Here, the decision-making processing performed by the information processing device 1A is, as an example, a hierarchical decision-making processing performed by multiple first derivation units 12-1, 121-2, ..., and second derivation unit 122, as described above.
[0076] The derived decision-making result is then executed to obtain a loss value corresponding to the next round. The loss value and a predicted value corresponding to the loss value are provided to the information processing device 1A and are referenced in the decision-making process in the next round. In this way, the information processing system 100A sequentially derives decision-making results.
[0077] (Specific Processing Example 1 by Information Processing Device 1A) Specific processing example 1 by information processing device 1A will be described below. Information processing device 1A according to this example executes the following algorithm 1 and algorithm 2. Algorithm 1 is mainly executed by each of the multiple first derivation units 121-1, 121-2, ..., and algorithm 2 is mainly executed by second derivation unit 122.
[0078] (Processing of Algorithm 1) First, the processing of Algorithm 1 will be described. As shown at the beginning of Algorithm 1, the acquisition unit 11 acquires the values of K, N, and η. Here, K is a first natural number that defines the total number of experts, and N is a second natural number that defines the total number of experts. Each expert Using two indexes i and n, it is specified two-dimensionally as (i, n), for example. Note that the index n can be considered as an index for introducing and identifying each so-called sleeping expert, but this does not limit this exemplary embodiment.
[0079] As will be described later, η is the first decision-making result (p t (i)) is derived from the weighting factor (with a tilde, w t (i, n) are parameters used in deriving
[0080] Next, as shown in “1:” of Algorithm 1, the derivation unit 12 calculates the reliability w of each expert in Round 1. 1 Set to 0.
[0081] Next, the control unit 10A executes the loop processing specified by "2:" to "12:" in algorithm 1. This loop processing is executed while incrementing an index t indicating the round until t = T (T is the total number of rounds). In other words, the control unit 10A repeats the processing specified by "3:" to "12:" in algorithm 1 for each round.
[0082] More specifically, as shown in “3:” of Algorithm 1, the acquisition unit 11 first acquires the predicted value m t where m t As shown in "3:" of Algorithm 1, for example, the predicted value m is obtained as a vector whose number of elements is the first natural number K. t is the loss l tHowever, this does not limit the present exemplary embodiment.
[0083] Next, as shown in “4:” of Algorithm 1, the first derivation unit 121 calculates the weighting factor (w t (i, n)) as the reliability of each expert w t (i, n), the predicted value m provided by each expert t (i), and using the parameter η Then, the first derivation unit 121 uses the weighting factor to derive the weighting factor as shown in "5:" of Algorithm 1. Therefore, the first decision-making result p t (i) is derived.
[0084] In this way, the first derivation unit 121 calculates each predicted value (m t ) and the reliability of each expert (w t ) and the first decision-making result (p t ) is derived.
[0085] As expressed by the above formula, each of the first derivation units 121-1, 121-2, ... calculates the reliability w t (i, n) and predicted value m t (i) and the weighting factor (with a tilde) t (i, n)) is calculated, and the first decision-making result (p t (i)) is derived.
[0086] As described above, the first derivation unit 121 derives the first decision-making result (p t Since the configuration for deriving (i) is adopted, when the number of experts increases, a weighting factor for the new expert can be added to the linear sum, and when the number of experts decreases, a weighting factor for the new expert can be deleted from the linear sum.
[0087] Therefore, this configuration makes it possible to realize a sequential decision-making process that can flexibly accommodate the addition or deletion of experts.
[0088] The derived first decision result p t is executed by the execution unit 21 of the terminal device 2A, as shown in "6:" of the algorithm 1. Then, the first decision-making result p t The loss value l corresponding to t is acquired by the loss value acquisition unit 22 of the terminal device 2A and provided to the information processing device 1A. t At least a part of the above may be supplied to a second derivation unit 122 that executes algorithm 2 described later without being executed.
[0089] In addition, the Δ K teeth, denotes the set defined by
[0090] Next, as shown in “7:” of Algorithm 1, the acquisition unit 11 calculates the reliability w t The parameter α to correct t Get.
[0091] Next, the loop process specified by "8:" to "12:" in Algorithm 1 is executed. The loop process is executed while incrementing the index n that designates each expert.
[0092] As shown in "9:" to "10:" in Algorithm 1, the condition t=d2 is expressed using an odd number d and an index n that specifies each expert. n If the above condition is satisfied, the first derivation unit 121 calculates the reliability w t+1 (i, n) is set to 0. The process shown in "9:" to "10:" in Algorithm 1 is to set the reliability w n Using a predetermined proportional coefficient d (d is a natural number) and a constant term g (g is an integer), t = d × 2 n This can also be expressed as an example of a process of resetting at timing (round) t expressed as +g.
[0093] In this way, in this example, as shown in "9:" to "10:" in Algorithm 1, the reliability of each of the multiple experts is reset at different times, thereby preventing a decrease in the accuracy of the decision-making results while also preventing an increase in computational costs.
[0094] On the other hand, as shown in “11:” to “12:” in Algorithm 1, if the above-mentioned condition regarding the number of rounds t is not satisfied, the first derivation unit 121 calculates the reliability of each expert as follows: Here, in the above formula, < , > indicates an inner product.
[0095] In this way, the first derivation unit 121 calculates each loss value (l t ) and calculate the confidence level (w t ) to update (w t+1 (Derive the following.)
[0096] As is clear from the above formula, the reliability is updated by the first derivation unit 121 using the correction parameter α t Therefore, the correction parameter α t By appropriately setting the reliability of each expert, it is possible to appropriately correct the reliability of each expert. For example, it is possible to assign a reliability of a predetermined level or higher to an expert who is more reliable than other experts. Therefore, with the above configuration, operation that guarantees a minimum level of accuracy, so-called conservative operation, is possible.
[0097] The correction parameter α t As mentioned above, can be expressed as a parameter for correcting the reliability, but as is clear from "12:" in Algorithm 1, the loss value l t (i) A parameter for correcting the predicted value m t It can also be considered as a parameter for correcting (i).
[0098] (Processing of Algorithm 2) Next, the processing of Algorithm 2 will be described. As shown at the beginning of Algorithm 2, the acquisition unit 11 acquires the values of K, N, and T. Here, K, N, and T have the same meaning as the corresponding parameters in Algorithm 1. Specifically, K is a first natural number that defines the total number of experts, N is a second natural number that defines the total number of experts, and T is the total number of rounds.
[0099] Furthermore, as indicated by "Input" at the beginning of Algorithm 2, Algorithm 2 is executed with reference to Base Algorithm B. Here, Base Algorithm B refers to Algorithm 1 described above, as an example. Algorithm 2 is executed with reference to a plurality of Base Algorithms B. In other words, the second derivation unit 122 executes the second algorithm in cooperation with a first algorithm that executes each of the plurality of Algorithms 1. Note that, in correspondence with Algorithm 1 being referred to as the base algorithm, Algorithm 2 may also be referred to as the master algorithm, but this name does not limit this exemplary embodiment.
[0100] Next, as shown in “1:” of Algorithm 2, the second derivation unit 122 uses the total number of rounds T to derive the parameter M as follows: where log 2 The symbols on both sides of T indicate ceiling functions. Therefore, the parameter M is calculated by the second derivation unit 122 using a real argument (log 2 T) is set to the smallest integer value (e.g., log 2 If T=3.3...then set M=4).
[0101] Here, the parameter M has the meaning of the total number of base algorithms (algorithm 1) that Algorithm 2 refers to, but this does not limit the present exemplary embodiment.
[0102] As shown in "1:" of Algorithm 2, the second derivation unit 122 calculates η(j) as Here, the index j is an index for distinguishing a plurality of base algorithms from one another, and takes an integer value from 1 to M, as shown by “1:” in Algorithm 2, for example.
[0103] Next, as shown in “2:” of Algorithm 2, the second derivation unit 122 calculates the weight factor w 1 ' is the parameter p 1 ' for each j Here, the weight factor w 1 ' is as shown in "2:" of Algorithm 2, where Δ M is the above-mentioned Δ K Here, we show the set obtained by replacing K with M.
[0104] Next, as shown in “3:” of Algorithm 2, the second derivation unit 122 calculates the Here, in the initialization process, as shown in “3:” of Algorithm 2, the second derivation unit 122 initializes the base algorithm B j This includes a process of passing the values of M, N, T, and η(j) to the
[0105] Next, the control unit 10A executes the loop processing specified by "4:" to "13:" in algorithm 2. This loop processing is executed while incrementing the index t indicating the round until t = T. In other words, the control unit 10A repeats the processing specified by "5:" to "13:" in algorithm 1 for each round.
[0106] More specifically, as shown in "5:" of Algorithm 2, the acquisition unit 11 first calculates the predicted value m t The second derivation unit 122 then transfers the predicted value m to each base algorithm (each algorithm 1) via each first derivation unit 121. t supply.
[0107] Next, as shown in "6:" of Algorithm 2, the acquisition unit 11 acquires the decision-making results p t,j Then, the second derivation unit 122 obtains the parameter h t (j) Here, < , > indicates the dot product.
[0108] Next, as shown in “7:” of Algorithm 2, the second derivation unit 122 The reliability vector w indicates the reliability of each base algorithm. t Here, D ψ teeth, where ψ denotes the Bregman divergence defined by is defined by
[0109] Next, as shown in “8:” of Algorithm 2, the second derivation unit 122 calculates the decision-making result p by each base algorithm (each Algorithm 1). t,j , a reliability vector w indicating the reliability of each base algorithm t Each component of the reliability w t (j) Using Therefore, the second decision result p t is derived.
[0110] In this way, the second derivation unit 122 calculates the first decision-making result (p t,j ) and the reliability (w t (j)) and the second decision-making result p t is derived.
[0111] More specifically, as expressed by the above formula, the second derivation unit 122 calculates the first decision-making results (p t,j ) and the reliability (w t (j)) is weighted according to the second decision-making result p tis derived.
[0112] As described above, the second derivation unit 122 calculates the first decision-making result (p t,j ) and the reliability (w t (j)) and the second decision-making result p t Here, the first decision-making result is made by referring to the predicted values provided by each expert as described above.
[0113] Therefore, with the above configuration, the first derivation unit 121 and the second derivation unit 122 can derive a suitable decision-making result by referring to the predicted values provided by the experts.
[0114] Next, as shown in “9:” of Algorithm 2, the derived second decision-making result p t is executed by the execution unit 21 of the terminal device 2A, for example. Then, the second decision-making result p t The loss value l corresponding to t is acquired by the loss value acquisition unit 22 of the terminal device 2A and provided to the information processing device 1A.
[0115] Next, as shown in “10:” of Algorithm 2, the acquisition unit 11 calculates the correction parameter α t Get.
[0116] Next, as shown in “11:” of Algorithm 2, the second derivation unit 122 calculates the loss value l obtained in “9:” of Algorithm 2. t and the correction parameter α obtained in "10:" of Algorithm 2. t Each component α t (j) is fed to the base algorithm Bj. The loss value l provided in this step is t As an example, the l obtained in "6:" of Algorithm 1 is t In addition, the α provided in this step corresponds to t (j) is the correction parameter α obtained in “7:” of Algorithm 1. t It corresponds to.
[0117] Next, as shown in “12:” of Algorithm 2, the acquisition unit 11 calculates the correction parameter β t Next, as shown in “13:” of Algorithm 2, the second derivation unit 122 obtains the weight factor w t ' is the weight factor w t+1 More specifically, the second derivation unit 122 updates the weight factor w t+1 'of, Here, g t teeth, is defined by b t teeth, Also, Δ K '(α) is is the set defined by
[0118] As described above, the second derivation unit 122 derives the loss value l and each of the plurality of first decision-making results p t,j With reference to the weight factor w t ', and the updated weight factor is the confidence w t+1 Therefore, according to the above configuration, it is possible to derive a decision-making result that is suitably adapted to a changeable environment.
[0119] As described above, the weighting factor w t ' is updated by the correction parameter β t Therefore, the correction parameter β t By appropriately setting the reliability of each of the plurality of first derivation units 122-j (in other words, the plurality of base algorithms), it is possible to appropriately correct the reliability of each of the plurality of first derivation units 122-j (in other words, the plurality of base algorithms). For example, it is possible to assign a predetermined or higher reliability to a base algorithm that is more reliable than other base algorithms. Therefore, the above configuration makes it possible to perform an operation that guarantees a minimum level of accuracy, that is, a so-called conservative operation.
[0120] The correction parameter β tcan be expressed as a parameter for correcting the reliability, but as is clear from "13:" in Algorithm 2, the loss value l t or the first decision-making result p t,j It can also be considered as a parameter for correcting the
[0121] As described above, the first derivation unit 121 and the second derivation unit 122, which execute Algorithm 1 and Algorithm 2, reset the reliability of each of the multiple experts constituting at least a portion of the multiple experts in each of the multiple Algorithms 1 (base algorithms) at mutually different times. Then, the second decision-making result is derived by referring to the first decision-making result obtained by such Algorithm 1 (base algorithm). Therefore, with the above configuration, it is possible to derive a suitable decision-making result while suppressing an increase in computational costs.
[0122] (Another Example of Base Algorithm) The base algorithm executed by the first derivation unit 121 is not limited to the above-mentioned algorithm 1. Below, we will explain algorithm 1', which is another example of the base algorithm. Algorithm 1' is mainly executed by each of the multiple first derivation units 121-1, 121-2, .... Furthermore, the second derivation unit 122 executes the above-mentioned algorithm 2 in cooperation with the multiple first derivation units 121 that execute algorithm 1'. As algorithm 2 has been described above, its description will be omitted. Algorithm 1' differs from Algorithm 1 in "9:", but is otherwise similar to Algorithm 1.
[0123] In Algorithm 1′, as shown in “9:” of Algorithm 1′, when the number of rounds t satisfies the condition t=dN+n expressed using a certain odd number d, an index n specifying each expert, and the second natural number N specifying the total number of experts, the first derivation unit 121 calculates the reliability w of the expert (i, n). t+1(i, n) is set to 0. The process shown in "9:" to "10:" in Algorithm 1' is to set the reliability w n This can also be expressed as an example of a process of resetting at timing (round) t expressed as t = d × N + n + g using a predetermined proportional coefficient d (d is a natural number) and a constant term g (g is an integer).
[0124] In this way, in this example, in Algorithm 1', the reliability of each of the multiple experts is reset at different times, as shown in "9:" to "10:" in Algorithm 1', so that it is possible to suppress a decrease in the accuracy of the decision-making results while suppressing an increase in calculation costs.
[0125] [Application Example] An application example in which the information processing system 100 according to this exemplary embodiment is applied to medicine and healthcare will be described below. Fig. 8 is a diagram schematically showing processing by the information processing device 1 according to this example.
[0126] 8, the information processing device 1 according to this example makes a decision regarding matching between a plurality of medical professionals and a plurality of hospitals (derives a decision-making result). Here, as described above, the information processing device 1 according to this example acquires, for a plurality of experts, predicted values and loss values associated with each expert in a certain round, and derives a decision-making result by referring to each acquired predicted value and each loss value.
[0127] The specific configuration of the information processing device 1 in this example does not limit this application example, but may be the same as the information processing device 1 described in exemplary embodiment 1, or may be the same as the information processing device 1A described in exemplary embodiment 2.
[0128] Furthermore, the decision-making process performed by the information processing device 1 in this example can be, as an example, a hierarchical decision-making process using multiple first derivation units 12-1, 121-2, ... and a second derivation unit 122, as described with respect to the information processing device 1A, but this does not limit this application example.
[0129] The derived decision-making result is then executed to obtain a loss value corresponding to the next round. The loss value and a predicted value corresponding to the loss value are provided to the information processing device 1 according to the present example and are referred to in the decision-making process in the next round.
[0130] Examples of inputs to each expert and outputs (predicted values) from each expert according to this example are as follows: As described in the first exemplary embodiment, the information processing system according to this example may include an "expert" or may acquire predicted values from an external "expert."
[0131] (Example of input to the expert) - Features of each hospital and each medical worker observed at time t (round t) - Number of patients visiting each hospital in the previous round (round t-1) (Example of output from the expert) - Hospital assigned to each medical worker at time t (round t).
[0132] (Processing Flow According to This Example) An example of processing by the information processing system 100 according to this example will be described below.
[0133] First, a person in charge of inputting information at the hospital inputs information (also referred to as hospital data) such as diagnosis status, availability of hospital rooms, medical departments, and consultation hours via a terminal device 2A or the like into the information processing system 100. The input information is stored in the storage unit 15A, for example, and referenced by the control unit 10A. The lower part of Fig. 9 shows an example of hospital data managed by the information processing system 100 according to this example.
[0134] Next, each medical worker inputs their own data (specialty, years of service, preferred hospital, etc.) (also referred to as medical worker data) into the information processing system 100. The input information is stored in the storage unit 15A, for example, and referenced by the control unit 10A. The upper part of Figure 9 shows an example of medical worker data managed by the information processing system 100 according to this example.
[0135] Next, the information processing device 1 according to this example derives a decision-making result regarding the optimal matching of hospitals and medical professionals by referring to the hospital data and the medical professional data. As an example, each of the multiple experts according to this example calculates a predicted value by referring to the hospital data and the medical professional data. Then, the information processing device 1 according to this example derives a decision-making result regarding the optimal matching of hospitals and medical professionals by referring to these predicted values.
[0136] The information processing system 100 according to the present example then proposes optimal hospital candidates to the medical worker via the terminal device 2A, etc. As an example, the information processing system 100 according to the present example presents optimal hospital candidates to the medical worker via a display panel, etc., provided in the terminal device 2A. Furthermore, the information processing system 100 according to the present example performs work-related registration for each medical worker.
[0137] For example, the information processing system 100 according to this example records the number of patients visiting each hospital each month (each round), and determines the assignment destination for the next round based on the loss value related to the number of patients visiting each hospital.
[0138] Here, the specific example of the loss value is not limited to this example, but as an example, a loss value according to the degree of congestion in a hospital can be used.
[0139] More specifically, the loss value can be calculated by (the actual number of patients visiting each hospital) - (the number of medical professionals assigned to each hospital x the number of patients that each medical professional can examine).
[0140] In addition, when referring to the predicted values by each expert or each first decision-making result, the information processing device 1 in this example may be configured to adopt the predicted values by each expert or each first decision-making result with a probability corresponding to the reliability, instead of using a weighted sum of these values according to the reliability.
[0141] As described above, the information processing system described in this specification can, for example, suitably make decisions regarding matching multiple medical professionals with multiple hospitals. Of course, application examples of the information processing system described in this specification are not limited to the above examples. The information processing system can be applied to various cases related to decision-making (prediction).
[0142] [Example of implementation by software] The control blocks (particularly the acquisition unit 11 and the derivation unit 12) of the information processing device 1, 1A and the terminal device 2, 2A may be implemented by a logic circuit (hardware) formed on an integrated circuit (IC chip) or the like, or may be implemented by software.
[0143] In the latter case, the information processing device 1, 1A and the terminal device 2, 2A each include a computer that executes instructions of a program, which is software that realizes each function. This computer includes, for example, at least one processor (control device) and at least one computer-readable recording medium that stores the program. The object of the present invention is achieved by having the processor in the computer read and execute the program from the recording medium.
[0144] The processor may be, for example, a CPU (Central Processing Unit), GPU (Graphics Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating point number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination thereof.
[0145] The recording medium may be a "non-transitory tangible medium," such as a ROM (Read Only Memory), a tape, a disk, a card, a semiconductor memory, or a programmable logic circuit. The computer may further include a RAM (Random Access Memory) for expanding the program. The program may be supplied to the computer via any transmission medium capable of transmitting the program (such as a communications network or broadcast waves). One aspect of the present invention may also be realized in the form of a data signal embedded in a carrier wave, in which the program is embodied by electronic transmission.
[0146] [Additional Note 1] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means disclosed in the above-described embodiments are also included in the technical scope of the present invention.
[0147] [Additional Note 2] Part or all of the above-described embodiment can also be described as follows: However, the present invention is not limited to the following described aspects.
[0148] (Supplementary Note 1) An information processing device comprising: an acquisition means for acquiring a predicted value (output value) obtained from each of a plurality of experts (models) and a loss value corresponding to each predicted value (output value); a derivation means (derivation unit 12) for deriving a decision-making result (optimal solution) according to each predicted value (output value) acquired by the acquisition means and the reliability of each expert (model); and an update means (derivation unit 12) for updating the reliability of each expert by referring to each loss value acquired by the acquisition means, wherein the update means resets (initializes) the reliability of each of a plurality of experts (models) constituting at least a part of the plurality of experts (models) at mutually different times.
[0149] (Supplementary Note 2) The information processing device according to Supplementary Note 1, wherein the derivation means further refers to a parameter for correcting reliability to derive the decision-making result (optimal solution).
[0150] (Supplementary Note 3) The updating means (deriving unit 12) calculates the reliability w of the n-th expert (model) among the plurality of experts (models). n Using a predetermined proportional coefficient d and constant term g, t = d × 2 n The information processing device according to claim 1 or 2, wherein the information processing device is reset (initialized) at a timing t represented as +g.
[0151] (Supplementary Note 4) The updating means (deriving unit 12) calculates the reliability w of the n-th expert (model) among the plurality of experts (models). n is reset (initialized) at a timing t expressed as t=d×N+n+g using a total number N of experts (models) and predetermined constants d and g.
[0152] (Supplementary Note 5) The information processing device described in any one of Supplementary Notes 1 to 4, wherein the derivation means comprises: a plurality of first derivation means that derive a first decision-making result (first optimal solution) according to each predicted value (output value) acquired by the acquisition means and the reliability of each expert (model); and a second derivation means that derives a second decision-making result (second optimal solution) according to the first decision-making result (first optimal solution) derived by each of the plurality of first derivation means and the reliability of each of the plurality of first derivation means.
[0153] (Supplementary Note 6) The information processing device according to Supplementary Note 5, wherein the first derivation means calculates a weighting factor for each expert (model) determined by the reliability and the predicted value (output value), and derives a first decision-making result (first optimal solution) using a linear sum of the weighting factors.
[0154] (Supplementary Note 7) The information processing device described in Supplementary Note 5 or 6, wherein the second derivation means derives the second decision-making result (second optimal solution) by a weighted sum of each of the first decision-making results (first optimal solutions) derived by the plurality of first derivation means, the weighted sum being in accordance with the reliability of each of the plurality of first derivation means.
[0155] (Supplementary Note 8) An information processing method including: an information processing device acquiring a predicted value (output value) obtained from each of a plurality of experts (models) and a loss value corresponding to each predicted value (output value); deriving a decision-making result (optimal solution) according to each of the acquired predicted values (output values) and the reliability of each expert (model), and updating the reliability of each expert (model) by referring to each of the acquired loss values; wherein in the updating step, the reliability of each of a plurality of experts (models) constituting at least a part of the plurality of experts (models) is reset (initialized) at mutually different times.
[0156] (Supplementary Note 9) A program that causes a computer to function as an information processing device, the program causing the computer to execute the following: acquiring a predicted value (output value) obtained from each of a plurality of experts (models) and a loss value corresponding to each predicted value (output value); deriving a decision-making result (optimal solution) according to each of the acquired predicted values (output values) and the reliability of each expert, and updating the reliability of each expert (model) by referring to each of the acquired loss values; and in the updating step, resetting (initializing) the reliability of each of a plurality of experts (models) that constitute at least a part of the plurality of experts (models) at mutually different times.
[0157] (Supplementary Note 10) An information processing system including an information processing device and a terminal device, wherein the information processing device comprises: an acquisition means for acquiring a predicted value (output value) obtained from each of a plurality of experts (models) and a loss value corresponding to each predicted value (output value); a derivation means for deriving a decision-making result (optimal solution) according to each predicted value (output value) acquired by the acquisition means and a reliability of each expert (model); and an update means for updating the reliability of each expert (model) by referring to each loss value acquired by the acquisition means, wherein the update means resets (initializes) the reliability of each of a plurality of experts (models) constituting at least a part of the plurality of experts (models) at mutually different times; and the terminal device comprises: an execution means for executing the decision-making result (optimal solution) derived by the information processing device; and a loss value acquisition means for acquiring a loss value obtained by executing the decision-making result (optimal solution).
[0158] [Additional Note 3] Part or all of the above-described embodiment can also be expressed as follows.
[0159] (Appendix A1) An information processing device comprising at least one processor, the processor executing: an acquisition process for acquiring output values obtained from each of a plurality of models and loss values corresponding to each output value; a derivation process for deriving an optimal solution according to each output value acquired in the acquisition process and the reliability of each model; and an update process for updating the reliability of each model by referring to each loss value acquired in the acquisition process, wherein in the update process, the reliability of each of a plurality of models constituting at least a part of the plurality of models is initialized at mutually different times.
[0160] (Supplementary Note A2) The information processing device according to Supplementary Note A1, wherein the processor derives the optimal solution by further referring to a parameter for correcting reliability in the derivation process.
[0161] (Appendix A3) In the update process, the processor nUsing a predetermined proportional coefficient d and constant term g, t = d × 2 n The information processing device according to Appendix A1 or A2, wherein the initialization is performed at a timing t represented as +g.
[0162] (Appendix A4) In the update process, the processor n The information processing device according to Appendix A1 or A2, wherein the information processing device initializes the model number n at a timing t expressed as t=d×N+n+g using a total number N of models and predetermined constants d and g.
[0163] (Appendix A5) The information processing device described in any one of Appendices A1 to A4, wherein the derivation process includes: a plurality of first derivation processes that derive a first optimal solution depending on each output value acquired in the acquisition process and the reliability of each model; and a second derivation process that derives a second optimal solution depending on the first optimal solution derived by each of the plurality of first derivation processes and the reliability of each of the plurality of first derivation processes.
[0164] (Supplementary Note A6) The information processing device according to Supplementary Note A5, wherein in the first derivation process, the processor calculates a weighting factor determined by the reliability and the output value for each model, and derives a first optimal solution using a linear sum of the weighting factors.
[0165] (Supplementary Note A7) The information processing device according to Supplementary Note A5 or A6, wherein in the second derivation process, the processor derives the second optimal solution by a weighted sum of each of the first optimal solutions derived in the plurality of first derivation processes, the weighted sum being in accordance with the reliability of each of the plurality of first derivation processes.
[0166] (Appendix A8) An information processing method, comprising: an information processing device acquiring output values obtained from each of a plurality of models and loss values corresponding to each output value; deriving an optimal solution according to each of the acquired output values and the reliability of each model; and updating the reliability for each model by referring to each of the acquired loss values, wherein in the updating step, the reliability of each of a plurality of models constituting at least a portion of the plurality of models is initialized at mutually different times.
[0167] (Appendix A9) A non-transitory recording medium storing a program that causes a computer to function as an information processing device, wherein the program causes the computer to: acquire output values obtained from each of a plurality of models and loss values corresponding to each output value; derive an optimal solution according to each of the acquired output values and the reliability of each model; and update the reliability of each model by referring to each of the acquired loss values, and in the updating step, initialize the reliability of each of a plurality of models that constitute at least a part of the plurality of models at mutually different times.
[0168] (Appendix A10) An information processing system including an information processing device and a terminal device, wherein the information processing device comprises one or more first processors, and the first processor executes: an acquisition process that acquires output values obtained from each of a plurality of models and loss values corresponding to each output value; a derivation process that derives an optimal solution according to each output value acquired in the acquisition process and a reliability of each model; and an update process that updates the reliability of each model by referring to each loss value acquired in the acquisition process, wherein the update process initializes the reliability of each of a plurality of models that constitute at least a part of the plurality of models at mutually different times; and the terminal device comprises one or more second processors, and the second processor executes: an execution process that executes the optimal solution derived by the information processing device; and a loss value acquisition process that acquires a loss value obtained by executing the optimal solution.
[0169] The information processing device and the information processing system may further include a memory that stores a program for causing the processor to execute the acquisition process, the derivation process, and the update process. The program may be recorded on a computer-readable, non-transitory, tangible recording medium.
[0170] REFERENCE SIGNS LIST 1, 1A... Information processing device 100, 100A... Information processing system 10A... Control unit 11... Acquisition unit 12... Derivation unit 121... First derivation unit 122... Second derivation unit S1, S100... Information processing method
Claims
1. An information processing device comprising: an acquisition means for acquiring an output value obtained from each of a plurality of models and a loss value corresponding to each output value; a derivation means for deriving an optimal solution based on each output value acquired by the acquisition means and the reliability of each model; and an update means for updating the reliability for each model by referring to each loss value acquired by the acquisition means, wherein the update means initializes the reliability of each of a plurality of models constituting at least a part of the plurality of models at mutually different times.
2. The information processing device according to claim 1, wherein said deriving means further refers to a parameter for correcting the reliability to derive said optimal solution.
3. The update means updates the reliability w of the n-th model among the plurality of models. n Using a predetermined proportional coefficient d and constant term g, t = d x 2 n The information processing apparatus according to claim 1 , wherein the initialization is performed at a timing t represented as +g.
4. The update means updates the reliability w of the n-th model among the plurality of models. n The information processing apparatus according to claim 1 , wherein n is initialized at a timing t expressed as t=d×N+n+g using a total number N of models and predetermined constants d and g.
5. An information processing device as claimed in any one of claims 1 to 4, wherein the derivation means comprises: a plurality of first derivation means for deriving a first optimal solution depending on each predicted value acquired by the acquisition means and the reliability of each model; and a second derivation means for deriving a second optimal solution depending on the first optimal solution derived by each of the plurality of first derivation means and the reliability of each of the plurality of first derivation means.
6. The information processing device according to claim 5, wherein the first derivation means calculates a weighting factor determined by the reliability and the output value for each model, and derives a first optimal solution using a linear sum of the weighting factors.
7. An information processing device as described in claim 5 or 6, wherein the second derivation means derives the second optimal solution by a weighted sum of the first optimal solutions derived by the plurality of first derivation means, the weighted sum being in accordance with the reliability of each of the plurality of first derivation means.
8. An information processing method comprising: an information processing device acquiring an output value obtained from each of a plurality of models and a loss value corresponding to each output value; deriving an optimal solution according to each of the acquired output values and the reliability of each model; and updating the reliability for each model by referring to each of the acquired loss values, wherein in the updating step, the reliability of each of a plurality of models constituting at least a portion of the plurality of models is initialized at mutually different times.
9. A program that causes a computer to function as an information processing device, the program causing the computer to execute the following: acquiring output values obtained from each of a plurality of models and loss values corresponding to each output value; deriving an optimal solution according to each of the acquired output values and the reliability of each model; and updating the reliability for each model by referring to each of the acquired loss values, wherein in the updating step, the reliability of each of a plurality of models constituting at least a portion of the plurality of models is initialized at mutually different times.
10. An information processing system including an information processing device and a terminal device, wherein the information processing device comprises: an acquisition means for acquiring an output value obtained from each of a plurality of models and a loss value corresponding to each output value; a derivation means for deriving an optimal solution based on each output value acquired by the acquisition means and a reliability of each model; and an update means for updating the reliability for each model by referring to each loss value acquired by the acquisition means, wherein the update means initializes the reliability of each of a plurality of models constituting at least a part of the plurality of models at mutually different times; and the terminal device comprises: an execution means for executing the optimal solution derived by the information processing device; and a loss value acquisition means for acquiring a loss value obtained by executing the optimal solution.
Citation Information
Patent Citations
Elevator traffic demand predicting system
JP2014172718A
Atmospheric pressure estimating device
JP2015203308A
Efficient load value prediction
US20200364055A1
