Information processing device, information processing method, and program

Abstract

This information processing device comprises at least one memory and at least one processor. The at least one processor acquires a latent variable on the basis of first weather data, converts the latent variable to a discrete expression, and calculates second weather data on the basis of the discrete expression.

Description

Information processing device, information processing method, and program

[0001] The present disclosure is the "Research and development of elemental technologies for providing user-optimized data using remote sensing technology" of the Ministry of Internal Affairs and Communications in fiscal year 2024, and relates to an information processing device, an information processing method, and a program.

[0002] For example, in the field of climate research or weather forecasting, numerical simulation is one of the important tools. A surrogate modeling method that speeds up numerical simulation based on deep learning is known (see, for example, Non-Patent Document 1).

[0003] L. Espeholt, et al., "Deep learning for twelve-hour precipitation forecasts," [online], [Retrieved December 20, 2024], Internet<URL: https: / / www.nature.com / articles / s41467-022-32483-x.pdf>

[0004] The present disclosure provides techniques for fast calculation of weather data.

[0005] An information processing device according to one aspect of the present disclosure has at least one memory and at least one processor, wherein the at least one processor obtains latent variables based on first weather data, converts the latent variables into a discrete representation, and calculates second weather data based on the discrete representation.

[0006] FIG. 1 is a block diagram showing an example of the overall configuration of a weather simulation system; FIG. 2 is a diagram for explaining an example of a prediction model; FIG. 3 is a flowchart showing an example of a learning process; FIG. 4 is a flowchart showing an example of a calculation process; FIG. 5 is a diagram showing an example of a relationship between a compression rate and a restoration performance; FIG. 6 is a diagram showing an example of a restoration result; FIG. 7 is a diagram showing an example of a simulation result; FIG. 8 is a block diagram showing an example of a hardware configuration of a computer.

[0007] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0008] In this specification, "^" is a symbol that should be written directly above the character immediately following it, but due to limitations in text notation, it is written immediately before it in the main text. In mathematical formulas, it is written directly above the original character.

[0009] An embodiment of the present disclosure may be an example of an information processing system that executes a weather simulation. The weather simulation may be a simulation of changes in atmospheric conditions, a simulation of changes in ocean conditions, or a simulation of a coupled atmosphere-ocean system. Hereinafter, the information processing system according to this embodiment may also be referred to as a "weather simulation system."

[0010] Numerical simulation has traditionally been an important tool in meteorological research and weather forecasting. However, high-dimensional meteorological simulation data requires large amounts of memory read / write, which necessitates large-scale computers. Furthermore, calculations that require large amounts of memory read / write are limited by memory bandwidth, which can prevent the full utilization of the computer's computing performance.

[0011] In recent years, surrogate modeling methods based on deep learning have been proposed to speed up numerical simulations. Some of these methods calculate the atmospheric state in a very high-dimensional three-dimensional space in a latent space obtained by nonlinear transformation. However, there is a limit to the compression rate of the dimensions of the latent space, because significantly compressing the dimensions of the latent space reduces the accuracy of restoring the atmospheric state. Therefore, conventional techniques have room for reducing memory capacity.

[0012] This embodiment may be intended to calculate meteorological data at high speed. In this embodiment, latent variables may be obtained based on first meteorological data, the latent variables may be converted into a discrete representation, and second meteorological data may be calculated based on the discrete representation. In one aspect, this embodiment calculates other meteorological data based on meteorological data compressed into a discrete representation, thereby enabling meteorological data to be calculated at high speed with less memory. In another aspect, this embodiment enables meteorological data to be calculated at high speed, allowing meteorological simulations to be performed with fine spatial meshes or time steps.

[0013] <Overall Configuration> The overall configuration of the weather simulation system in this embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the overall configuration of the weather simulation system.

[0014] 1 , the weather simulation system 1000 may include a simulation device 10. The weather simulation system 1000 may include a terminal device that inputs data to the simulation device 10 and displays data output from the simulation device 10.

[0015] The simulation device 10 may be an example of an information processing device that executes a weather simulation, and may be a computer such as a personal computer, a workstation, or a server.

[0016] Weather data to be used as a target for a weather simulation may be input to the simulation device 10. Hereinafter, the weather data to be used as a target for a weather simulation will also be referred to as "target data." The target data is an example of first weather data.

[0017] The target data may be meteorological data that serves as the initial values ​​for the weather simulation. The target data may be data generated by the weather simulation. The target data may be data generated by assimilating observed values ​​obtained by observing actual weather information. For example, the observed values ​​may be observed using a temperature sensor, a humidity sensor, a barometric pressure sensor, an anemometer, a meteorological satellite, or a weather radar installed in the area that is the target of the weather simulation.

[0018] The weather data may include one or more weather-related variables. The weather-related variables may include variables indicating atmospheric conditions or ocean conditions. The weather-related variables may include, for example, at least one of wind speed (east-west, north-south, vertical), temperature, air pressure, humidity (water vapor mixing ratio), cloud water, rain, ice, snow, or hail (mixing ratio).

[0019] The weather data may indicate weather information for each position information in a three-dimensional space. The position information may include, for example, altitude information, latitude information, and longitude information. The weather data may be, for example, four-dimensional data with dimensions of variables: altitude, latitude, and longitude. The weather data may be multidimensional data including multiple four-dimensional data for each variable. The weather data may be multidimensional data including multiple variables for each combination of altitude, latitude, and longitude.

[0020] The simulation device 10 may output meteorological data indicating the calculation results of the meteorological simulation. Hereinafter, the meteorological data indicating the calculation results will also be referred to as "result data." The result data is an example of second meteorological data.

[0021] The simulation device 10 may include a prediction model M. The prediction model M may be a machine learning model that receives target data as input and outputs result data. The prediction model M may include, for example, a neural network, a vector quantized variational autoencoder (VQ-VAE), a convolutional neural network (CNN), a neural network with an attention mechanism, or the like. The neural network with an attention mechanism may be, for example, a neural network called a Transformer.

[0022] The simulation device 10 may learn a prediction model M. The simulation device 10 may learn a plurality of prediction models M for each weather-related variable. The simulation device 10 may learn the prediction model M based on learning data. The learning data may be weather data collected in advance. For example, the learning data may include observation data obtained by observing actual weather information, weather data obtained by assimilating weather forecast data provided from a weather information system or the like, or weather data including simulation results generated by a weather simulation.

[0023] The learning data may be stored in advance in a storage device. For example, the storage device in which the learning data is stored may be a memory of the simulation device 10 or an external storage device that can communicate with the simulation device 10 via a communication network.

[0024] The overall configuration of the weather simulation system 1000 shown in FIG. 1 is an example, and various system configuration examples are possible depending on the application and purpose. The weather simulation system 1000 may be configured with one or more devices. The devices included in the weather simulation system 1000 may be a system configured with multiple devices. Each function included in the weather simulation system 1000 may be realized by any device that constitutes the system. Each component included in the weather simulation system 1000 may be included in any device that constitutes the system.

[0025] For example, a plurality of simulation devices 10 may be included in the weather simulation system 1000. The simulation device 10 may be realized by a plurality of computers, or may be realized as a cloud computing service. The classification of devices such as the simulation device 10 shown in FIG. 1 is one example.

[0026] <Prediction Model> The prediction model of the simulation device 10 will be described with reference to Fig. 2. Fig. 2 is a diagram showing an example of the prediction model.

[0027] As shown in FIG. 2 , the prediction model M may include a generation model M1 and a surrogate model M2. The prediction model M may receive target data as input and output result data. In this embodiment, the target data and result data are four-dimensional data with the dimensions of the variable, altitude, latitude, and longitude. That is, in this embodiment, a prediction model M that predicts a single variable will be described. However, the prediction model M may also predict multiple variables simultaneously.

[0028] The generative model M1 may include an encoding unit E, a quantization unit Q, a decoding unit D, and a codebook C. As an example, the generative model M1 may be a vector quantization variational autoencoder (VQ-VAE). As an example, the encoding unit E, the quantization unit Q, and the decoding unit D may each be configured by a convolutional neural network (CNN) or a transformer. As an example, the encoding unit E and the decoding unit D may each be configured by a three-dimensional U-Net type network structure.

[0029] The encoding unit E may receive the target data input to the generative model M1. In this embodiment, the target data is x∈R Ａｌｔ×Ｌａｔ×Ｌｏｎ×Ｖａｒ where R is a real number, Alt is altitude, Lat is latitude, Lon is longitude, and Var is a variable.

[0030] The encoding unit E may encode the target data x. The encoding unit E may acquire a latent variable Z based on the target data x. The encoding unit E may acquire the latent variable Z by compressing the target data x in the spatial direction. The encoding unit E may compress the target data x based on a predetermined resolution f. The encoding unit E may compress the target data x by 1 / f times in the spatial direction. The resolution f may be, for example, 2 ０～５ (i.e., 1, 2, 4, 8, 16, 32). The resolution f may be specified by the user. In this embodiment, the latent variable Z is Z∈R Ａｌｔ／ｆ×Ｌａｔ／ｆ×Ｌｏｎ／ｆ×Ｄ The latent variable Z is defined as the latent variable z∈R Ｄ where D is the number of dimensions of the latent variable.

[0031] The encoding unit E may output the latent variable Z obtained from the target data x. The latent variable Z output from the encoding unit E may be input to the quantization unit Q.

[0032] The quantization unit Q may quantize the latent variable Z. The quantization unit Q quantizes the latent variable Z into a discrete representation z ｅ (x) can be converted into the discrete representation z ｅ (x) may be a natural number between 1 and K, inclusive, where K may be any natural number. For example, K may be between 512 and 8192, inclusive.

[0033] The codebook C is a set of K latent variables e ｋ ∈R Ｄ where k is a natural number between 1 and K. The codebook C may include a codebook C for each predetermined resolution f. The codebook C may include one codebook C for multiple resolutions f. ｅ The compression ratio to (x) is determined by the number of resolutions f and the number of codebooks C.

[0034] The quantization unit Q converts the latent variable z into a discrete representation z based on the codebook C. ｅ The quantization unit Q may convert the latent variable z to the latent variable e in the codebook C. ｋ The quantization unit Q may search for the latent variable e closest to the latent variable z. ｋ The discrete representation z of the latent variable z ｅ (x). As an example, the quantizer Q may calculate the equation (1).

[0035]

[0036] However, ||| ２ is the L2 norm of the value .

[0037] The quantization unit Q quantizes the latent variable z into a discrete representation z ｅ The discrete representation z output from the quantization unit Q may be ｅ (x) may be input to the decoder D and the surrogate model M2.

[0038] The decoder D may restore the target data x. The decoder D may restore the discrete representation z ｅ The decoder D may reconstruct the target data x based on the discrete representation z ｅ The target data x may be restored by decoding (x). The decoding unit D may output the target data ^x obtained as a restoration result. The target data ^x is an example of third weather data.

[0039] The generative model M1 may be trained based on a predetermined loss function. The predetermined loss function may be configured to reduce the difference between the target data x input to the encoding unit E and the target data ^x output from the decoding unit D. The predetermined loss function may also be configured to reduce the difference between the latent variable z before quantization and the discrete representation z after quantization. ｅ (x) may be configured to reduce the difference between (x) and (x). As an example, the loss function may be a function Loss defined by equation (2).

[0040]

[0041] where sg[·] is the stop gradient operator. The generative model M1 may use dynamic initialization if memory capacity allows. Using dynamic initialization can improve the codebook utilization rate.

[0042] The surrogate model M2 is a discrete representation z output from the quantizer Q. ｅ The surrogate model M2 may generate result data based on the discrete representation z ｅ The result data may be generated by evolving (x) over time. The surrogate model M2 may generate result data indicating weather information after a predetermined time has elapsed from the date and time of the target data. The predetermined time may be, for example, about six hours. In other words, the surrogate model M2 may generate result data indicating the simulation results of weather information six hours later.

[0043] The surrogate model M2 may be a machine learning model that has previously learned time evolution in a discrete space. The surrogate model M2 may include an image recognition unit and a decoding unit. For example, the image recognition unit may be configured with a Vision Transformer or a Swin Transformer. For example, the decoding unit may be configured with a convolutional neural network or a transformer.

[0044] Both the generative model M1 and the surrogate model M2 may be optimized simultaneously. For example, if the simulation device 10 has sufficient computational resources (e.g., a GPU, etc.), the generative model M1 and the surrogate model M2 may be optimized simultaneously.

[0045] The generative model M1 may include a regularization term to promote learning of a surrogate model M2 that performs future predictions based on the discrete representation obtained by the generative model M1. As an example, input data x at time t may be expressed as x ｔ Let x ｔ is coded by the coding unit E and then expressed as z ｔ Let x be the input data at time t+1. ｔ＋１ Let x ｔ＋１ is coded by the coding unit E and then expressed as z ｔ＋１ If the time between time t and time t+1 is short, x ｔ and x ｔ＋１ In this case, z ｔ and Z ｔ＋１ Therefore, the generative model M1 is ｔ and Z ｔ＋１ The regularization term may include a regularization term that reduces the distance between . As an example, the regularization term may be a cosine similarity.

[0046] The function Loss defined in equation (2) avoids the occurrence of non-differentiable operations due to discretization by introducing the gradient stopping operator sg[·]. Here, the gradient stopping operator sg[·] may be approximated by a continuous function. ｅ(x) is calculated by equation (1), but may be approximated by equation (3).

[0047]

[0048] where T is a positive constant. In the limit where T is infinite, the nearest e ｊ Only weight lol ｊ = 1 and other weights w are zero, so the discrete expression z calculated by Equation (3) ｅ (x) coincides with the calculation result of equation (1).

[0049] <Flow of Weather Simulation Method> The weather simulation method executed by the weather simulation system 1000 will be described with reference to Figures 3 and 4. The weather simulation method may include a learning process (see Figure 3) and a calculation process (Figure 4).

[0050] 3 is a flowchart showing an example of the learning process, which is an example of an information processing method for learning the prediction model M.

[0051] In step S1, the simulation device 10 acquires learning data. The simulation device 10 may read out the learning data stored in a storage device.

[0052] In step S2, the simulation device 10 divides the training data acquired in step S1, and acquires a plurality of partial data obtained by the division.

[0053] The simulation device 10 may divide the training data by space. The simulation device 10 may divide the training data into a plurality of partial data based on location information included in the training data. The simulation device 10 may divide the training data into a plurality of partial data based on altitude information. The simulation device 10 may divide the training data into a plurality of partial data based on at least one of latitude information and longitude information.

[0054] The simulation device 10 may divide the training data on a variable-by-variable basis. The simulation device 10 may divide the training data including multiple variables into multiple partial data sets each including one or more variables. As an example, the simulation device 10 may divide the training data including n variables into n partial data sets each including one variable. The simulation device 10 may divide the training data including n variables into n / k partial data sets each including k variables. Here, n is any natural number, and k is a natural number less than n.

[0055] In step S3, the simulation device 10 acquires latent variables based on each of the partial data divided in step S2. Specifically, the simulation device 10 inputs the partial data to the prediction model M. The partial data input to the prediction model M is input to the encoding unit E of the generation model M1.

[0056] The encoding unit E acquires latent variables based on the input partial data. The encoding unit E may compress the partial data in the spatial direction based on a predetermined resolution f. This generates latent variables in which the partial data is compressed by 1 / f times in the spatial direction. The encoding unit E outputs the acquired latent variables. The latent variables output from the encoding unit E are input to the quantization unit Q.

[0057] In step S4, the simulation device 10 quantizes the latent variables acquired in step S3. Specifically, the quantization unit Q of the generative model M1 converts the input latent variables into discrete representations. The quantization unit Q may convert the latent variables into discrete representations based on a codebook C corresponding to a predetermined resolution f. This generates discrete representations in which the latent variables are quantized. The quantization unit Q outputs the generated discrete representations. The discrete representations output from the quantization unit Q are input to the decoding unit D.

[0058] In step S5, the simulation device 10 restores the training data based on the discrete representation generated in step S4. Specifically, the decoding unit D of the generative model M1 restores the partial data input to the generative model M1 in step S3 based on the input discrete representation. The decoding unit D outputs the restored partial data. The simulation device 10 may restore the training data acquired in step S1 by combining the partial data output from the decoding unit D.

[0059] In step S6, the simulation device 10 updates the parameters of the generative model M1. Specifically, the simulation device 10 updates the parameters of the generative model M1 so as to minimize the loss function Loss shown in equation (2). The simulation device 10 may update the model parameters of the encoding unit E, the model parameters of the quantization unit Q, the model parameters of the decoding unit D, and the codebook C.

[0060] In step S7, the simulation device 10 determines whether or not learning of the prediction model M is complete. Whether or not learning is complete may be determined based on whether or not the parameters of the generation model M1 have converged, whether or not a predetermined number of iterations has been exceeded, or the like.

[0061] If the learning is completed (YES), the simulation device 10 proceeds to step S8. On the other hand, if the learning is not completed (NO), the simulation device 10 returns to step S1. Thereafter, the simulation device 10 acquires unlearned learning data in step S1 and executes the processes from step S2 to step S7 again. The simulation device 10 repeatedly executes the processes from step S1 to step S7 until it is determined in step S7 that the learning is completed.

[0062] In step S8, the simulation device 10 stores the trained prediction model M in the storage device. The simulation device 10 may store the trained prediction model M in the memory of the simulation device 10. The simulation device 10 may store the trained prediction model M in an external storage device that can communicate with the simulation device 10 via a communication network.

[0063] 4 is a flowchart illustrating an example of the calculation process, which is an example of an information processing method for executing a weather simulation using the trained prediction model M.

[0064] In step S11, the simulation device 10 acquires target data. The simulation device 10 may read the target data stored in a storage device. The simulation device 10 may acquire result data generated in a calculation process executed in the past as the target data.

[0065] In step S12, the simulation device 10 divides the target data acquired in step S11. The simulation device 10 acquires a plurality of partial data obtained by the division. The simulation device 10 may divide the observation data by space unit. The simulation device 10 may divide the observation data by variable unit.

[0066] In step S13, the simulation device 10 acquires latent variables based on each of the partial data divided in step S12. Specifically, the simulation device 10 inputs the partial data to the prediction model M. The partial data input to the prediction model M is input to the encoding unit E of the generation model M1.

[0067] The encoding unit E acquires latent variables based on the input partial data. The encoding unit E outputs the acquired latent variables. The latent variables output from the encoding unit E are input to the quantization unit Q.

[0068] In step S14, the simulation device 10 quantizes the latent variables acquired in step S13. The quantization unit Q outputs a discrete representation generated by the quantization. The discrete representation output from the quantization unit Q is input to the surrogate model M2.

[0069] In step S15, the simulation device 10 generates result data based on the discrete representation generated in step S14. Specifically, the surrogate model M2 evolves the input discrete representation over time to generate result data indicating weather information after a predetermined time has elapsed. The surrogate model M2 outputs the generated result data.

[0070] The simulation device 10 may execute the processes from step S13 to step S15 in parallel for each of the plurality of partial data divided in step S12. By dividing the meteorological data into a plurality of partial data in predetermined processing units and executing a simulation for each partial data in parallel, it is possible to execute a meteorological simulation covering a vast area (for example, the entire globe) at high speed.

[0071] In step S16, the simulation device 10 acquires the result data generated in step S15. Specifically, the simulation device 10 acquires the result data output from the surrogate model M2. The simulation device 10 outputs the acquired result data.

[0072] The simulation device 10 may store the result data in a storage device. For example, the simulation device 10 may store the result data in a memory of the simulation device 10. The simulation device 10 may store the result data in an external storage device that can communicate with the simulation device 10 via a communication network. The simulation device 10 may display the result data on a display device. The simulation device 10 may transmit the result data to an external information processing device or information processing system that can communicate with the simulation device 10 via a communication network.

[0073] <Evaluation Results> The evaluation results of the weather simulation system 1000 will be described with reference to Fig. 5 to Fig. 7. In this evaluation, the relationship between compression ratio and restoration performance (see Fig. 5), restoration performance (see Fig. 6), and simulation performance (see Fig. 7) were evaluated.

[0074] This evaluation used a dataset containing weather forecast data. Specifically, a dataset containing three hours of weather forecast data for each of three days in the summers of 2019 and 2021 was used. The target area was the Kanto region of Japan. The weather forecast data was divided into 256 x 256 grids with 500-meter intervals on the ground surface, and vertically divided into 23 layers from an altitude of 55 m to approximately 16 km. The time interval for variables was 30 seconds. Eleven variables were used: wind speed (east-west, north-south, vertical) [m / v], temperature [°C], atmospheric pressure [hPa], humidity (water vapor mixing ratio), water cloud, ice cloud, rain, snow, and hail.

[0075] This dataset was sampled at 10-minute intervals, and ensembles were created with 1, 10, and 10 members for each of the three days, with the average for the last two days added together to create a total of 4,892 samples. Of the 4,892 samples, 90% were used for training, and the remaining 10% were used for evaluation. Additionally, as a preprocessing step, the samples were standardized by variable and altitude.

[0076] FIG. 5 shows an example of the relationship between compression ratio and restoration performance. The compression ratio is determined by file size. For example, when the resolution f=8 and the number of codebooks K=8192, the compression ratio is 13863.3 times. The evaluation index used is the peak signal-to-noise ratio (PSNR). Specifically, the PSNR for the maximum fluctuation was used for each variable and altitude. The larger the PSNR value, the smaller the error (i.e., the higher the evaluation).

[0077] In Fig. 5, the horizontal axis represents the compression ratio (logarithmic scale) and the vertical axis represents the restoration performance (PSNR), and the restoration performance at each compression ratio is plotted. As shown in Fig. 5, the lower the compression ratio, the higher the restoration performance, and the higher the compression ratio, the lower the restoration performance. In other words, it can be seen that there is a trade-off between the compression ratio and the restoration performance.

[0078] FIG. 6 shows an example of a restoration result. The restoration result is the output of generative model M1 (i.e., the output of decoder D) when target data x is input. FIG. 6 shows observed data (true values) for wind speed (east-west, north-south, vertical), temperature, air pressure, and humidity, as well as the output results of two prediction models with different parameters (Proposed Method 1 and Proposed Method 2). Proposed Method 1 shows the output result when a prediction model predicting a single variable is configured with a compression rate of 20,000. Proposed Method 2 shows the output result when a prediction model predicting multiple variables is configured with a compression rate of 1,316. ​​As shown in FIG. 6, it can be seen that the prediction model predicting a single variable can be restored with high accuracy even at a compression rate of 20,000. It can also be seen that the prediction model predicting multiple variables can be restored with high accuracy at a compression rate of 1,316.

[0079] FIG. 7 is a diagram illustrating an example of a simulation result. The simulation result is the output of the prediction model M (i.e., the output of the surrogate model M2) when target data x is input. FIG. 7 illustrates the observed data (true values) after 30 minutes, the output results of the prediction model (proposed method), and the output results of the continuous model for the same six variables as in FIG. 6. The proposed method is the same as proposed method 2 shown in FIG. 6 (i.e., a compression rate of 1316 times). The continuous model is a conventional surrogate model (i.e., uncompressed) trained in continuous space. As shown in FIG. 7, the proposed method can perform simulations with the same accuracy as the uncompressed continuous model, even at a compression rate of 1316 times.

[0080] <Summary> As is clear from the above description, the simulation device 10 according to an embodiment of the present disclosure obtains latent variables based on first weather data, converts the latent variables into discrete representations, and calculates second weather data based on the discrete representations.

[0081] The first weather data may include weather information at a first time point. The second weather data may include weather information at a second time point after the first time point. The simulation device 10 may calculate the second weather data by evolving the discrete representation over time.

[0082] The simulation device 10 may divide the first meteorological data into a plurality of partial data, and may obtain a latent variable for each of the partial data.

[0083] The meteorological data may include a variable indicating meteorological information for each piece of spatial position information. The simulation device 10 may divide the first meteorological data into a plurality of partial data pieces based on the position information.

[0084] The location information may include altitude information in space. The simulation device 10 may divide the first meteorological data into a plurality of partial data based on the altitude information.

[0085] The location information may include latitude information and longitude information in space. The simulation device 10 may divide the first meteorological data into a plurality of partial data based on at least one of the latitude information and the longitude information.

[0086] The meteorological data may include a plurality of variables indicating a plurality of types of meteorological information. The simulation device 10 may divide the first meteorological data into partial data including one or more variables.

[0087] The simulation device 10 may acquire latent variables by compressing the first meteorological data in a spatial direction. The simulation device 10 may compress the first meteorological data in a spatial direction based on a predetermined resolution. The simulation device 10 may convert the latent variables into a discrete representation based on the predetermined resolution.

[0088] The simulation device 10 may convert the latent variables into discrete representations based on a trained codebook. The codebook may be trained so that the third weather data reconstructed based on the discrete representations approaches the first weather data.

[0089] According to an embodiment of the present disclosure, other weather data is calculated based on weather data compressed into a discrete representation, so that weather data can be calculated at high speed. In one aspect, according to this embodiment, since weather data can be calculated at high speed, weather simulation can be performed with fine spatial meshes or time steps.

[0090] [Hardware Configuration of Information Processing Device] Some or all of the devices (simulation device 10) in the above-described embodiments may be configured with hardware, or may be configured with software (program) information processing executed by a CPU (Central Processing Unit), GPU (Graphics Processing Unit), or the like. When configured with software information processing, software that realizes at least some of the functions of each device in the above-described embodiments may be stored on a non-transitory storage medium (non-transitory computer-readable medium) such as a CD-ROM (Compact Disc-Read Only Memory) or a USB (Universal Serial Bus) memory, and the software information processing may be executed by loading the software into a computer. The software may also be downloaded via a communication network. Furthermore, all or part of the software processing may be implemented in a circuit such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array), thereby allowing the software information processing to be executed by hardware.

[0091] The storage medium that stores the software may be a removable medium such as an optical disk, or a fixed medium such as a hard disk, memory, etc. The storage medium may be provided inside the computer (such as a main storage device or auxiliary storage device) or outside the computer.

[0092] 8 is a block diagram showing an example of the hardware configuration of each device (simulation device 10) in the above-described embodiment. Each device may be realized as a computer 7 including, for example, a processor 71, a main storage device 72 (memory), an auxiliary storage device 73 (memory), a network interface 74, and a device interface 75, which are connected via a bus 76.

[0093] Although the computer 7 in FIG. 8 includes one of each component, it may also include multiple of the same component. Also, while FIG. 8 shows one computer 7, the software may be installed on multiple computers, and each of the multiple computers may execute the same or different parts of the software. In this case, a distributed computing configuration may be used in which each computer communicates with the other computers via a network interface 74 or the like to execute the processing. In other words, each device (simulation device 10) in the above-described embodiment may be configured as a system in which one or more computers execute instructions stored in one or more storage devices to realize its functions. Furthermore, it may also be configured such that information transmitted from a terminal is processed by one or more computers provided on a cloud, and the processing results are transmitted to the terminal.

[0094] The various calculations of each device (simulation device 10) in the above-described embodiments may be executed in parallel using one or more processors, or using multiple computers via a network. Furthermore, the various calculations may be distributed to multiple processing cores within a processor and executed in parallel. Furthermore, some or all of the processes, means, etc. disclosed herein may be implemented by at least one of a processor and a storage device provided on a cloud that can communicate with the computer 7 via a network. Thus, each device in the above-described embodiments may be implemented in the form of parallel computing using one or more computers.

[0095] The processor 71 may be an electronic circuit (processing circuit, processing circuitry, CPU, GPU, FPGA, ASIC, etc.) that at least controls or performs calculations on a computer. The processor 71 may be a general-purpose processor, a dedicated processing circuit designed to perform a specific calculation, or a semiconductor device that includes both a general-purpose processor and a dedicated processing circuit. The processor 71 may also include an optical circuit or a calculation function based on quantum computing.

[0096] The processor 71 may perform arithmetic processing based on data or software input from each device or the like configured internally of the computer 7, and may output arithmetic results or control signals to each device or the like. The processor 71 may control each component constituting the computer 7 by executing the OS (Operating System) of the computer 7, applications, etc.

[0097] Each apparatus (simulation apparatus 10) in the above-described embodiment may be realized by one or more processors 71. Here, the processor 71 may refer to one or more electronic circuits arranged on one chip, or may refer to one or more electronic circuits arranged on two or more chips or two or more devices. When multiple electronic circuits are used, the electronic circuits may communicate with each other via wire or wirelessly.

[0098] The main memory device 72 may store instructions executed by the processor 71, various data, etc., and information stored in the main memory device 72 may be read by the processor 71. The auxiliary memory device 73 is a memory device other than the main memory device 72. Note that these memory devices refer to any electronic component capable of storing electronic information and may be semiconductor memory. The semiconductor memory may be either volatile memory or non-volatile memory. The memory device for saving various data, etc. in each device (simulation device 10) in the above-described embodiment may be realized by the main memory device 72 or the auxiliary memory device 73, or may be realized by an internal memory built into the processor 71. For example, each memory unit in the above-described embodiment may be realized by the main memory device 72 or the auxiliary memory device 73.

[0099] When each device (simulation device 10) in the above-described embodiment is configured with at least one storage device (memory) and at least one processor connected (coupled) to this at least one storage device, at least one processor may be connected to one storage device. Also, at least one storage device may be connected to one processor. Also, a configuration in which at least one processor among multiple processors is connected to at least one storage device among multiple storage devices may be included. Also, this configuration may be realized by storage devices and processors included in multiple computers. Furthermore, a configuration in which a storage device is integrated with a processor (for example, a cache memory including an L1 cache and an L2 cache) may be included.

[0100] The network interface 74 is an interface for connecting to the communication network 8 wirelessly or via a wire. The network interface 74 may be an appropriate interface, such as one that conforms to an existing communication standard. The network interface 74 may exchange information with an external device 9A connected via the communication network 8. The communication network 8 may be any one of a wide area network (WAN), a local area network (LAN), a personal area network (PAN), etc., or a combination thereof, as long as information is exchanged between the computer 7 and the external device 9A. An example of a WAN is the Internet, an example of a LAN is IEEE 802.11 or Ethernet (registered trademark), and an example of a PAN is Bluetooth (registered trademark) or NFC (Near Field Communication), etc.

[0101] The device interface 75 is an interface such as a USB that directly connects to the external device 9B.

[0102] The external device 9A is a device connected to the computer 7 via a network, and the external device 9B is a device connected directly to the computer 7.

[0103] The external device 9A or the external device 9B may be, for example, an input device. The input device is, for example, a camera, a microphone, a motion capture device, various sensors, a keyboard, a mouse, a touch panel, or the like, and provides acquired information to the computer 7. Alternatively, the external device 9A or the external device 9B may be a device including an input unit, a memory, and a processor, such as a personal computer, a tablet terminal, or a smartphone.

[0104] Furthermore, the external device 9A or the external device 9B may be, for example, an output device. The output device may be, for example, a display device such as an LCD (Liquid Crystal Display) or an organic EL (Electro Luminescence) panel, or a speaker that outputs sound or the like. Alternatively, the external device 9A or the external device 9B may be a device including an output unit, a memory, and a processor, such as a personal computer, a tablet terminal, or a smartphone.

[0105] The external device 9A or the external device 9B may be a storage device (memory). For example, the external device 9A may be a network storage or the like, and the external device 9B may be a storage device such as an HDD.

[0106] Furthermore, the external device 9A or the external device 9B may be a device having some of the functions of the components of each device (simulation device 10) in the above-described embodiment. That is, the computer 7 may transmit some or all of the processing results to the external device 9A or the external device 9B, or may receive some or all of the processing results from the external device 9A or the external device 9B.

[0107] In this specification (including the claims), when the expression "at least one of a, b, and c" or "at least one of a, b, or c" (including similar expressions) is used, it includes any of a, b, c, ab, ac, bc, or abc. It may also include multiple instances of any element, such as aa, abb, aabbcc, etc. Furthermore, it also includes the addition of elements other than the enumerated elements (a, b, and c), such as having d, as in abcd.

[0108] In this specification (including claims), when expressions such as "using / using data as input / based on / according to / in response to data" (including similar expressions) are used, unless otherwise specified, this includes cases where the data itself is used, or where data that has been processed in some way (e.g., data with noise added, normalized data, features extracted from data, intermediate representations of data, etc.) is used. Furthermore, when a statement is made that a result is obtained "using data as input / based on / according to / in response to data" (including similar expressions), this includes cases where the result is obtained based solely on the data, or where the result is influenced by other data, factors, conditions, and / or states other than the data. Furthermore, when a statement is made that "data is output" (including similar expressions), this includes cases where the data itself is used as output, or where data that has been processed in some way (e.g., data with noise added, normalized data, features extracted from data, intermediate representations of various data, etc.) is used as output, unless otherwise specified.

[0109] When the terms "connected" and "coupled" are used in this specification (including the claims), they are intended as open-ended terms that include any of direct connection / coupling, indirect connection / coupling, electrically connection / coupling, communicatively connection / coupling, functionally connection / coupling, and physically connection / coupling. These terms should be interpreted appropriately according to the context in which they are used, but any connection / coupling form that is not intentionally or naturally excluded should be interpreted as being included in these terms without limitation.

[0110] In this specification (including the claims), the expression "A configured to B" may include the physical structure of element A having a configuration capable of performing operation B, and the permanent or temporary setting / configuration of element A being configured / set to actually perform operation B. For example, if element A is a general-purpose processor, it is sufficient that the processor has a hardware configuration capable of performing operation B, and is configured to actually perform operation B by setting a permanent or temporary program (instruction). Also, if element A is a dedicated processor, dedicated arithmetic circuit, etc., it is sufficient that the circuit structure of the processor is implemented to actually perform operation B, regardless of whether control instructions and data are actually attached.

[0111] Whenever words implying containing or possessing (e.g., "comprising / including," "having," etc.) are used in this specification (including the claims), they are intended to be open-ended terms that include containing or possessing things other than the object designated by the object of the term. When the object of such words implying containing or possessing does not specify a quantity or suggests a singular number (e.g., expressions using the articles "a" or "an"), the expression should be construed as not being limited to a specific number.

[0112] In this specification (including the claims), although expressions such as "one or more" and "at least one" are used in some places and expressions that do not specify a quantity or that imply a singular number (expressions using the articles "a" or "an") are used in other places, the latter expressions are not intended to mean "one." In general, expressions that do not specify a quantity or that imply a singular number (expressions using the articles "a" or "an") should be interpreted as not necessarily being limited to a specific number.

[0113] In this specification, when a particular advantage / result is described as being obtained with respect to a particular configuration of an embodiment, it should be understood that the same advantage / result can also be obtained with one or more other embodiments having the same configuration, unless otherwise stated. However, it should be understood that the presence or absence of the effect generally depends on various factors, conditions, and / or situations, and that the effect is not necessarily obtained with the configuration. The effect is merely obtained by the configuration described in the embodiment when various factors, conditions, and / or situations are satisfied, and the effect does not necessarily occur in a claimed invention that defines the same configuration or a similar configuration.

[0114] In this specification (including claims), when multiple pieces of hardware perform a predetermined process, the pieces of hardware may cooperate to perform the predetermined process, or some of the hardware may perform all of the predetermined process. Furthermore, some of the hardware may perform part of the predetermined process, and other hardware may perform the rest of the predetermined process. In this specification (including claims), when an expression such as "one or more pieces of hardware perform a first process, and the one or more pieces of hardware perform a second process" (including similar expressions) is used, the hardware performing the first process and the hardware performing the second process may be the same or different. In other words, it is sufficient that the hardware performing the first process and the hardware performing the second process are included in the one or more pieces of hardware. Note that hardware may include an electronic circuit, a device including an electronic circuit, etc.

[0115] In this specification (including the claims), when multiple storage devices (memories) store data, each of the multiple storage devices may store only a portion of the data, or may store the entire data. Also, a configuration in which only some of the multiple storage devices store data may be included.

[0116] In this specification (including the claims), terms such as "first," "second," etc. are used merely as a way of distinguishing between two or more elements, and are not necessarily intended to impose technical meanings such as temporal aspect, spatial aspect, sequence, quantity, etc. on the subject. Thus, for example, a reference to a first element and a second element does not necessarily mean that only two elements may be employed therein, that the first element must precede the second element, that the first element must be present in order for the second element to be present, etc.

[0117] Although the embodiments of the present disclosure have been described in detail above, the present disclosure is not limited to the individual embodiments described above. Various additions, modifications, substitutions, partial deletions, etc. are possible within the scope of the conceptual idea and spirit of the present invention, which is derived from the content defined in the claims and their equivalents. For example, when numerical values ​​or formulas are used in the above-described embodiments, they are shown for illustrative purposes and do not limit the scope of the present disclosure. Furthermore, the order of each operation shown in the embodiments is also illustrative and does not limit the scope of the present disclosure.

[0118] The disclosed technology may take the following forms as described below.

[0119] (Supplementary Note 1) An information processing device having at least one memory and at least one processor, wherein the at least one processor obtains latent variables based on first weather data, converts the latent variables into a discrete representation, and calculates second weather data based on the discrete representation.

[0120] (Supplementary Note 2) The information processing device according to Supplementary Note 1, wherein the first weather data includes weather information at a first time point, and the second weather data includes weather information at a second time point that is later than the first time point.

[0121] (Supplementary Note 3) The information processing device according to Supplementary Note 2, wherein the at least one processor calculates the second weather data by evolving the discrete representation over time.

[0122] (Supplementary Note 4) The information processing device according to any one of Supplementary Notes 1 to 3, wherein the at least one processor divides the first meteorological data into a plurality of partial data, and acquires the latent variable for each of the partial data.

[0123] (Supplementary Note 5) The information processing device according to Supplementary Note 4, wherein the weather data includes variables indicating weather information for each spatial position information, and the at least one processor divides the first weather data into a plurality of partial data based on the position information.

[0124] (Supplementary Note 6) The information processing device according to Supplementary Note 5, wherein the location information includes altitude information in the space, and the at least one processor divides the first meteorological data into a plurality of partial data based on the altitude information.

[0125] (Supplementary Note 7) The information processing device according to Supplementary Note 5, wherein the location information includes latitude information and longitude information in the space, and the at least one processor divides the first meteorological data into a plurality of partial data based on at least one of the latitude information and the longitude information.

[0126] (Supplementary Note 8) The information processing device according to Supplementary Note 4, wherein the weather data includes a plurality of variables indicating a plurality of types of weather information, and the at least one processor divides the first weather data into the partial data including one or more of the variables.

[0127] (Supplementary Note 9) The information processing device according to any one of Supplementary Notes 1 to 8, wherein the at least one processor acquires the latent variables by compressing the first meteorological data in a spatial direction.

[0128] (Supplementary Note 10) The information processing device according to Supplementary Note 9, wherein the at least one processor compresses the first meteorological data in a spatial direction based on a predetermined resolution, and converts the latent variables into the discrete representation based on the resolution.

[0129] (Supplementary Note 11) The information processing device according to any one of Supplementary Notes 1 to 10, wherein the at least one processor converts the latent variables into the discrete representation based on a trained codebook.

[0130] (Supplementary Note 12) The information processing device according to Supplementary Note 11, wherein the codebook is trained so that third weather data reconstructed based on the discrete representation approaches the first weather data.

[0131] (Supplementary Note 13) An information processing method, comprising: at least one processor: obtaining latent variables based on first weather data; converting the latent variables into a discrete representation; and calculating second weather data based on the discrete representation.

[0132] (Supplementary Note 14) A program for causing at least one processor to execute the following process: acquiring latent variables based on first meteorological data; converting the latent variables into a discrete representation; and calculating second meteorological data based on the discrete representation.

[0133] This application claims priority to Provisional Application No. 63 / 671,934, filed July 16, 2024, in the United States Patent and Trademark Office, the entire contents of which are incorporated herein by reference.

[0134] M: Prediction model M1: Generative model M2: Surrogate model E: Encoding unit Q: Quantization unit D: Decoding unit C: Codebook 10: Simulation device 1000: Weather simulation system

Claims

1. An information processing device having at least one memory and at least one processor, wherein the at least one processor obtains latent variables based on first weather data, converts the latent variables into a discrete representation, and calculates second weather data based on the discrete representation.

2. The information processing device according to claim 1, wherein the first weather data includes weather information at a first time point, and the second weather data includes weather information at a second time point after the first time point.

3. The information processing device according to claim 2, wherein the at least one processor calculates the second meteorological data by evolving the discrete representation over time.

4. The information processing device according to any one of claims 1 to 3, wherein the at least one processor divides the first meteorological data into a plurality of partial data, and acquires the latent variables for each of the partial data.

5. The information processing device described in claim 4, wherein the weather data includes variables indicating weather information for each spatial position information, and the at least one processor divides the first weather data into multiple partial data based on the position information.

6. The information processing device according to claim 5, wherein the location information includes altitude information in the space, and the at least one processor divides the first meteorological data into a plurality of partial data based on the altitude information.

7. The information processing device of claim 5, wherein the location information includes latitude information and longitude information in the space, and the at least one processor divides the first meteorological data into a plurality of partial data based on at least one of the latitude information or the longitude information.

8. The information processing device according to claim 4, wherein the weather data includes a plurality of variables indicating a plurality of types of weather information, and the at least one processor divides the first weather data into the partial data including one or more of the variables.

9. The information processing device according to any one of claims 1 to 8, wherein the at least one processor acquires the latent variables by compressing the first meteorological data in a spatial direction.

10. The information processing device according to claim 9, wherein the at least one processor compresses the first meteorological data in a spatial direction based on a predetermined resolution, and converts the latent variables into the discrete representation based on the resolution.

11. The information processing device according to any one of claims 1 to 10, wherein the at least one processor converts the latent variables into the discrete representation based on a trained codebook.

12. The information processing device according to claim 11, wherein the codebook is trained so that the third weather data reconstructed based on the discrete representation approaches the first weather data.

13. An information processing method, comprising: at least one processor: obtaining latent variables based on first weather data; converting the latent variables into a discrete representation; and calculating second weather data based on the discrete representation.

14. A program for causing at least one processor to execute the following process: obtaining latent variables based on first meteorological data; converting the latent variables into a discrete representation; and calculating second meteorological data based on the discrete representation.

Images