Analysis device, method, and program
The analysis device and method efficiently solve optimal transportation problems by block diagonalizing the cost matrix and using unitary transformations, enabling fast computation of optimal transportation matrices and distances without approximation.
Patent Information
- Application Number
- PCT/JP2023/020353
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-31
- Publication Date
- 2024-12-05
AI Technical Summary
Solving the optimal transportation problem at high speed without approximation is challenging due to its computational intensity, despite existing methods approximating it to simpler forms.
The solution involves an analysis device and method that calculates a cost matrix with Nth-order cyclic symmetry, block diagonalizes it, and uses a unitary matrix to convert probability vectors, reducing the problem size, allowing for quick solving of optimal transportation problems by outputting an optimal transportation matrix and distance.
This approach enables the rapid solution of optimal transportation problems without approximation, significantly reducing computational time by transforming the problem into a smaller, equivalent form.
Smart Images

Figure JP2023020353_05122024_PF_FP_ABST
Abstract
Description
Analytical device, method and program
[0001] FIELD Embodiments of the present invention relate to an analytical device, a method, and a program.
[0002] The optimal transportation problem is a problem of finding a transportation route that minimizes the total cost when transporting the cost matrix from one probability vector to the other probability vector, given two probability vectors and a cost matrix between them (see, for example, non-patent documents 1 and 2).
[0003] Marco Cuturi. Sinkhorn distances: Lightspeed computation of optimal transport. In Advances in Neural Information Processing Systems, 2013. Evgeny Tenetov, Gershon Wolansky, and Ron Kimmel. Fast entropic regularized optimal transport using semidiscrete cost approximation. SIAM Journal on Scientific Computing, 40(5):A3400-A3422, 2018.
[0004] We now define the mathematical notation we will be working with: The sets of integers, real numbers, and complex numbers are represented as follows:
[0005]
[0006] Positive odd numbers are expressed as follows:
[0007]
[0008] A set can also be expressed as follows:
[0009]
[0010] In terms of this set, it can be expressed as follows:
[0011]
[0012] When other inequalities are used in this notation, they similarly mean that all elements of the set satisfy the condition of that inequality.
[0013] Also, in terms of matrices, it can be expressed as follows:
[0014]
[0015] Also, all elements of the matrix are expressed as follows:
[0016]
[0017] The transpose, conjugate transpose, real part and imaginary part of the matrix are expressed as follows:
[0018]
[0019] For a d-dimensional probability simplex, it can be expressed as follows:
[0020]
[0021] The Frobenius norm of a matrix is expressed as follows:
[0022]
[0023] For a d-dimensional vector, it can be expressed as follows:
[0024]
[0025] We define the discrete optimal transportation problem to be addressed. First, for two d-dimensional probability vectors, we express them as follows:
[0026]
[0027] Furthermore, a cost matrix of d×d dimensions is expressed as follows:
[0028]
[0029] Given the above two d-dimensional probability vectors and a d×d-dimensional cost matrix, the optimal transportation problem is defined as a constrained minimization problem of the following equation (1).
[0030]
[0031] Here, for this optimal transportation problem, the transportation matrix is expressed as follows:
[0032]
[0033] The optimal transportation problem is defined as follows:
[0034]
[0035] In addition to the optimal transport matrix, the optimal value d of this objective function is called the optimal transport distance, and is used in various application programs.
[0036] It is known that solving the optimal transportation problem expressed as above takes a very long time. While the above-mentioned Non-Patent Documents 1 and 2 describe methods for solving the problem quickly by approximating it to a form that is easier to solve, there are still no promising methods for solving the problem quickly without approximation. This invention was made in light of the above circumstances, and its purpose is to provide an analytical device, method, and program that can solve the optimal transportation problem quickly without approximation.
[0037] An analysis device according to one embodiment of the present invention comprises a cost matrix calculation unit that calculates a single cost matrix obtained by contracting a cost matrix having N-th-order cyclic symmetry; a probability vector calculation unit that calculates a probability vector obtained by transforming two probability vectors using a unitary matrix; an optimal transportation problem solving unit that outputs an optimal transportation matrix and an optimal transportation distance by solving an optimal transportation problem that is smaller in size than the original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit; and an original optimal transportation matrix output unit that uses the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is the solution to the original optimal transportation problem.
[0038] An analysis device according to one embodiment of the present invention includes a cost matrix block diagonalization unit that outputs N block cost matrices by block diagonalizing a cost matrix having N-th-order cyclic symmetry, where N is a power of 2; a cost matrix calculation unit that calculates a single cost matrix by contracting the N block cost matrices output by the cost matrix block diagonalization unit; a probability vector calculation unit that calculates a probability vector obtained by transforming two probability vectors using a unitary matrix; an optimal transportation problem solving unit that outputs an optimal transportation matrix and an optimal transportation distance by solving an optimal transportation problem that is smaller in size than the original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit; and an original optimal transportation matrix output unit that uses the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is a solution to the original optimal transportation problem.
[0039] An analysis method according to one embodiment of the present invention is a method performed by an analysis device, and includes the following steps: a cost matrix calculation unit of the analysis device calculates a single cost matrix by contracting a cost matrix having N-th-order cyclic symmetry; a probability vector calculation unit of the analysis device calculates a probability vector by transforming two probability vectors using a unitary matrix; an optimal transportation problem solving unit of the analysis device solves an optimal transportation problem that is smaller in size than the original optimal transportation problem, and is composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit, thereby outputting an optimal transportation matrix and an optimal transportation distance; and an original optimal transportation matrix output unit of the analysis device uses the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is a solution to the original optimal transportation problem.
[0040] An analysis method according to one aspect of the present invention is a method performed by an analysis device, the method comprising: a cost matrix block diagonalization unit of the analysis device block-diagonalizing a cost matrix having N-th-order cyclic symmetry, where N is a power of two, to output N block cost matrices; a cost matrix calculation unit of the analysis device calculating a single cost matrix by contracting the N block cost matrices output by the cost matrix block diagonalization unit; a probability vector calculation unit of the analysis device calculating a probability vector by transforming two probability vectors using a unitary matrix; an optimal transportation problem solving unit of the analysis device solving an optimal transportation problem that is smaller in size than the original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit, to output an optimal transportation matrix and an optimal transportation distance; and an original optimal transportation matrix output unit of the analysis device using the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix that is a solution to the original optimal transportation problem.
[0041] According to the present invention, the optimal transportation problem can be solved quickly without approximation.
[0042] Fig. 1 is a diagram showing an application example of an analysis device according to a first embodiment of the present invention. Fig. 2 is a diagram showing an example of a procedure for a processing operation by the analysis device according to the first embodiment of the present invention. Fig. 3 is a diagram showing an application example of an analysis device according to a second embodiment of the present invention. Fig. 4 is a diagram showing an example of a procedure for a processing operation by the analysis device according to the second embodiment of the present invention. Fig. 5 is a block diagram showing an example of the hardware configuration of an analysis device according to an embodiment of the present invention.
[0043] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In this embodiment, it is assumed that input data has N-th order cyclic symmetry, and a given optimal transportation problem is converted into an identical but smaller optimal transportation problem, i.e., the number of variables handled is 1 / N. 2 By converting the optimal transportation problem into one that is twice as simple, it becomes possible to solve the optimal transportation problem quickly.
[0044] First Embodiment Next, a first embodiment will be described. In this first embodiment, an optimization algorithm that can quickly solve an optimal transportation problem is described, assuming that input data has N-th order cyclic symmetry and further assuming that N is a power of 2.
[0045] The fact that N is a power of 2 can be expressed as follows:
[0046]
[0047] The optimization algorithm in the first embodiment is an optimization algorithm that can realize each of the following units (a) to (e): (a) a cost matrix block diagonalization unit that block diagonalizes an input cost matrix and outputs N block cost matrices; (b) a small cost matrix calculation unit that uses the N block cost matrices output from the cost matrix block diagonalization unit to calculate and output a single small cost matrix that condenses the information of the N block cost matrices; (c) a small probability vector calculation unit that converts two input probability vectors into smaller ones and outputs them; (d) a small optimal transportation problem solving unit that solves a small optimal transportation problem composed of the small cost matrix output from the small cost matrix calculation unit and the small probability vector output from the small probability vector calculation unit, and outputs a small optimal transportation matrix and an optimal transportation distance; (e) an original optimal transportation matrix output unit that uses the small optimal transportation matrix output from the small optimal transportation problem solving unit to find and output an optimal transportation matrix that is a solution to the original optimal transportation problem.
[0048] Fig. 1 is a diagram showing an application example of an analysis device according to a first embodiment of the present invention. Fig. 2 is a diagram showing an example of a procedure of processing operations by the analysis device according to the first embodiment of the present invention. As shown in Fig. 1, the analysis device 100 according to the first embodiment of the present invention includes a cost matrix block diagonalization unit 11, a small cost matrix calculation unit 12, a small probability vector calculation unit 13, a small optimal transportation problem solving unit 14, and an original optimal transportation matrix output unit 15.
[0049] The individual components are described below. The cost matrix block diagonalization unit 11 outputs a block cost matrix by block-diagonalizing the cost matrix using a unitary matrix obtained using, for example, the algorithm proposed in "Albert Fassler and Eduard Stiefel. Group Theoretical Methods and Their Applications. Birkhauser, Boston, 1992," under the assumption that the cost matrix input in the above formula (1) has N-th order cyclic symmetry and N is a power of 2 (S11).
[0050] The above unitary matrix is expressed as follows:
[0051]
[0052] The output block cost matrix is expressed as follows:
[0053]
[0054] The block diagonalization is performed according to the following equation (2).
[0055]
[0056] The small cost matrix calculation unit 12 calculates and outputs a single small cost matrix by condensing the information using the block cost matrix output from the cost matrix block diagonalization unit 11 (S12). This small cost matrix is expressed as follows:
[0057]
[0058] First, the small cost matrix calculation unit 12 solves the recurrence formula shown in the following equation (3) for p=1, . . . , P−1 in order, and for n=0, .
[0059]
[0060] However, the above formula (3) can be expressed as the following formula (4).
[0061]
[0062] Then, the small cost matrix calculation unit 12 calculates and outputs a small cost matrix according to the following equation (5).
[0063]
[0064] However, the above formula (5) can be expressed as the following formula (6).
[0065]
[0066] Based on the two probability vectors input in the above formula (1), the small probability vector calculation unit 13 calculates and outputs a small probability vector that is smaller in size than the original probability vector as follows (S13). This small probability vector is expressed as follows:
[0067]
[0068] Furthermore, this small probability vector is expressed as in the following equation (7).
[0069]
[0070] However, the notation of the above formula (7) is expressed as follows:
[0071]
[0072] The small optimal transportation problem solving unit 14 solves the small optimal transportation problem of the following equation (8), which is composed of the small cost matrix output from the small cost matrix calculation unit 12 and the small probability vector output from the small probability vector calculation unit 13, and outputs the small optimal transportation matrix and the optimal transportation distance d (S14).
[0073]
[0074] The small optimal transport matrix is expressed as follows:
[0075]
[0076] Note that d in the above formula (1) and formula (8) has a completely equivalent value. In this embodiment, the algorithm for solving this small optimal transport problem is not particularly limited. As an example of an algorithm, the above formula (8) can be solved by using the network simplex method given in "Gabriel Peyr'e and Marco Cuturi. Computational optimal transport. arXiv, 2018."
[0077] The original optimal transport matrix output unit 15 uses the sub-optimal transport matrix output by the sub-optimal transport problem solving unit 14 to find and output the optimal transport matrix that is the solution to the original optimal transport problem (S15). As with the cost matrix, the optimal transport matrix is block-diagonalized using the same unitary matrix as in equation (2). Therefore, a relational expression such as the following equation (9) exists.
[0078]
[0079] Here, the small optimal transport matrix satisfies the following equation, and the block optimal transport matrix (hereinafter sometimes referred to as a parameter) on the left side of this equation has already been found.
[0080]
[0081] Then, the original optimal transport matrix output unit 15 performs the calculation of the following equation (10).
[0082]
[0083] However, the notation of the above formula (10) is expressed as follows:
[0084]
[0085] Then, the original optimal transport matrix output unit 15 performs the calculation of the following equation (11).
[0086]
[0087] The original optimal transport matrix output unit 15 converts the following equation (12) into pBy solving for p=P-1, . . . , 2, the left side of equation (12) can be found for n=1, . . . , N / 2-1.
[0088]
[0089] However, the notation of the above formula (12) is expressed as follows:
[0090]
[0091] Then, the original optimal transport matrix output unit 15 obtains the left side of the following equation (13) by solving the equation (13) for n=1, . . . , N / 2−1.
[0092]
[0093] The original optimal transport matrix output unit 15 was able to determine the parameter on the left side of the following equation (hereinafter referred to as parameter n) for n=0, . . . , N / 2 from the above equations (8), (10) to (13).
[0094]
[0095] If the input data has cyclic symmetry, the following relationship holds:
[0096]
[0097] Therefore, the original optimal transport matrix output unit 15 can obtain all of the parameters n on the left side of the equation above for n = 0, ..., N-1. Therefore, the original optimal transport matrix output unit 15 can obtain and output the original optimal transport matrix from the above equation (9).
[0098] Second Embodiment Next, a second embodiment will be described. In this second embodiment, an optimization algorithm that can quickly solve an optimal transportation problem by assuming that input data has N-th order cyclic symmetry will be described.
[0099] The optimization algorithm in the second embodiment is an optimization algorithm that can realize each of the following units (a) to (d): (a) a small cost matrix calculation unit that uses an input cost matrix to calculate and output a small cost matrix that condenses the information of the input cost matrix; (b) a small probability vector calculation unit that converts two input probability vectors into small ones and outputs them; (c) a small optimal transportation problem solving unit that solves a small optimal transportation problem consisting of the small cost matrix that is the output of the small cost matrix calculation unit and the small probability vector that is the output of the small probability vector calculation unit, and outputs a small optimal transportation matrix and an optimal transportation distance; (d) an original optimal transportation matrix output unit that uses the small optimal transportation matrix that is the output of the small optimal transportation problem solving unit to find and output an optimal transportation matrix that is a solution to the original optimal transportation problem.
[0100] Fig. 3 is a diagram showing an application example of the analysis device according to the second embodiment of the present invention, and Fig. 4 is a diagram showing an example of a procedure of a processing operation by the analysis device according to the second embodiment of the present invention.
[0101] As shown in FIG. 3, an analysis device 100a according to the second embodiment of the present invention includes a small cost matrix calculation unit 12a, a small probability vector calculation unit 13a, a small optimal transportation problem solving unit 14a, and an original optimal transportation matrix output unit 15a.
[0102] The individual components are described below. In the small cost matrix calculation unit 12a, assuming that the cost matrix given in the above formula (1) has N-th order cyclic symmetry, the cost matrix can be expressed as a block circulant matrix as shown in the following formula (14).
[0103]
[0104] At this time, the small cost matrix calculation unit 12a calculates and outputs one small cost matrix of dimensions (d / N) × (d / N) by condensing the cost matrix information using the following equation (15) (S21).
[0105]
[0106] However, the notation of the above formula (15) is expressed as follows:
[0107]
[0108] Next, the small probability vector calculation unit 13a performs the same processing as the small probability vector calculation unit 13 in the first embodiment and outputs a small probability vector (S22). Next, the small optimal transportation problem solving unit 14a performs the same processing as the small optimal transportation problem solving unit 14 in the first embodiment and outputs a small optimal transportation matrix (S23). Note, however, that the small cost matrix used in the processing is expressed by the above formula (15).
[0109] Next, the original optimal transport matrix output unit 15a uses the sub-optimal transport matrix output by the sub-optimal transport problem solving unit 14a to find and output an optimal transport matrix that is a solution to the original optimal transport problem (S24). Similar to the cost matrix, the transport matrix is a block circulant matrix, so the optimal transport matrix can be expressed as in the following equation (16).
[0110]
[0111] However, the notation of the above formula (16) is expressed as follows:
[0112]
[0113] At this time, the following equation (17) holds true.
[0114]
[0115] However, the above formula (17) is defined as the following formula (18).
[0116]
[0117] However, the notation of the above formula (18) is expressed as follows:
[0118]
[0119] The original optimal transport matrix output unit 15a performs the calculation using the above equation (17) for n = 0, ..., N-1, thereby determining and outputting the original optimal transport matrix using the above equation (16).
[0120] Fig. 5 is a block diagram showing an example of the hardware configuration of an analytical device according to one embodiment of the present invention. In the example shown in Fig. 5, the analytical device 100 according to the first embodiment is configured, for example, by a server computer or a personal computer, and has a hardware processor 111A such as a CPU (Central Processing Unit). A program memory 111B, a data memory 112, an input / output interface 113, and a communication interface 114 are connected to this hardware processor 111A via a bus 115. The same applies to the analytical device 100a according to the second embodiment.
[0121] The communication interface 114 includes, for example, one or more wireless communication interface units, and enables transmission and reception of information to and from a communication network NW. As the wireless interface, for example, an interface that adopts a low-power wireless data communication standard such as a wireless LAN (Local Area Network) is used.
[0122] An input device 300 and an output device 400, which are attached to the analysis apparatus 100 and used by a user or the like, are connected to the input / output interface 113. The input / output interface 113 receives operation data input by a user or the like through the input device 300, such as a keyboard, touch panel, touchpad, or mouse, and outputs output data to an output device 400, which may include a display device using a liquid crystal or organic electroluminescence (EL) display, for display. The input device 300 and the output device 400 may be devices built into the analysis apparatus 100, or may be input devices and output devices of other information terminals that can communicate with the analysis apparatus 100 via the network NW.
[0123] The program memory 111B is a non-transitory tangible storage medium that is a combination of a non-volatile memory that can be written to and read from at any time, such as a hard disk drive (HDD) or a solid state drive (SSD), and a non-volatile memory such as a read only memory (ROM), and stores programs necessary to execute various control processes, etc., according to one embodiment.
[0124] The data memory 112 is a tangible storage medium that is, for example, a combination of the above-mentioned nonvolatile memory and a volatile memory such as RAM (Random Access Memory), and is used to store various data acquired and created during various processes performed by the analysis device 100.
[0125] The analytical device 100 according to one embodiment of the present invention may be configured as a data processing device having software-based processing function units. Storage areas used as work memory or the like by each unit of the analytical device 100 may be configured using the data memory 112 shown in FIG. 5. However, these configured storage areas are not essential components within the analytical device 100, and may be areas provided in, for example, an external storage medium such as a USB (Universal Serial Bus) memory, or a storage device such as a database server located in the cloud.
[0126] The processing function unit can be realized by having the hardware processor 111A read and execute a program stored in the program memory 111B, but the processing function unit may also be realized in various other forms, including an integrated circuit such as an application specific integrated circuit (ASIC) or a field-programmable gate array (FPGA).
[0127] The methods described in each embodiment may be stored as a program (software means) that can be executed by a computer on a recording medium such as a magnetic disk (e.g., a floppy disk, a hard disk, etc.), an optical disk (e.g., a CD-ROM, a DVD, an MO, etc.), or a semiconductor memory (e.g., a ROM, a RAM, a flash memory, etc.), or may be transmitted and distributed via a communication medium. The program stored on the medium also includes a configuration program that configures the software means (including not only executable programs but also tables and data structures) that the computer executes. The computer that realizes this device reads the program stored on the recording medium and, in some cases, configures the software means using the configuration program, and executes the above-described processing by having the operation controlled by this software means. The term "recording medium" as used herein is not limited to a storage medium for distribution, but also includes a storage medium such as a magnetic disk or semiconductor memory installed inside the computer or in a device connected via a network.
[0128] The present invention is not limited to the above-described embodiments, and various modifications can be made in the implementation stage without departing from the spirit of the invention. Furthermore, the embodiments may be implemented in appropriate combinations, in which case the combined effects can be obtained. Furthermore, the above-described embodiments include various inventions, and various inventions can be extracted by combining selected elements from the disclosed elements. For example, if the problem can be solved and the desired effect can be obtained even if some elements are deleted from all elements shown in the embodiments, the configuration from which these elements are deleted can be extracted as an invention.
[0129] 100, 100a... Analysis device 11... Cost matrix block diagonalization unit 12, 12a... Small cost matrix calculation unit 13, 13a... Small probability vector calculation unit 14, 14a... Small optimal transportation problem solving unit 15, 15a... Original optimal transportation matrix output unit
Claims
1. An analysis device comprising: a cost matrix calculation unit that calculates a single cost matrix obtained by condensing a cost matrix having N-th order cyclic symmetry; a probability vector calculation unit that calculates a probability vector obtained by transforming two probability vectors using a unitary matrix; an optimal transportation problem solving unit that outputs an optimal transportation matrix and an optimal transportation distance by solving an optimal transportation problem that is smaller in size than the original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit; and an original optimal transportation matrix output unit that uses the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is the solution to the original optimal transportation problem.
2. The analysis device according to claim 1, wherein the cost matrix calculation unit calculates the single contracted cost matrix based on a minimum value for each element of a block circulant matrix corresponding to the cost matrix.
3. The analysis device described in claim 1, wherein the original optimal transport matrix output unit outputs the original optimal transport matrix based on the optimal transport matrix output by the optimal transport problem solving unit and an index related to the minimum value for each component of a block circulant matrix corresponding to the cost matrix.
4. An analysis device comprising: a cost matrix block diagonalization unit that outputs N block cost matrices by block diagonalizing a cost matrix having N-th order cyclic symmetry, where N is a power of 2; a cost matrix calculation unit that calculates one cost matrix obtained by contracting the N block cost matrices output by the cost matrix block diagonalization unit; a probability vector calculation unit that calculates a probability vector obtained by transforming two probability vectors using a unitary matrix; an optimal transportation problem solving unit that outputs an optimal transportation matrix and an optimal transportation distance by solving an optimal transportation problem that is smaller in size than an original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit; and an original optimal transportation matrix output unit that uses the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is a solution to the original optimal transportation problem.
5. A method performed by an analysis device, comprising: a cost matrix calculation unit of the analysis device calculating a single cost matrix obtained by contracting a cost matrix having N-th order cyclic symmetry; a probability vector calculation unit of the analysis device calculating a probability vector obtained by converting two probability vectors using a unitary matrix; an optimal transportation problem solving unit of the analysis device solving an optimal transportation problem that is smaller in size than the original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit, thereby outputting an optimal transportation matrix and an optimal transportation distance; and an original optimal transportation matrix output unit of the analysis device using the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is a solution to the original optimal transportation problem.
6. A method performed by an analysis device, comprising: a cost matrix block diagonalization unit of the analysis device block diagonalizing a cost matrix having N-th order cyclic symmetry, where N is a power of 2, to output N block cost matrices; a cost matrix calculation unit of the analysis device calculating a single cost matrix by contracting the N block cost matrices output by the cost matrix block diagonalization unit; a probability vector calculation unit of the analysis device calculating a probability vector in which two probability vectors are transformed using a unitary matrix; an optimal transportation problem solving unit of the analysis device solving an optimal transportation problem that is smaller in size than the original optimal transportation problem, the optimal transportation problem being composed of the cost matrix calculated by the cost matrix calculation unit and the probability vector calculated by the probability vector calculation unit, to output an optimal transportation matrix and an optimal transportation distance; and an original optimal transportation matrix output unit of the analysis device using the optimal transportation matrix output by the optimal transportation problem solving unit to output the original optimal transportation matrix, which is a solution to the original optimal transportation problem.
7. An analytical processing program that causes a processor to function as each part of the analytical device according to any one of claims 1 to 4.
Citation Information
Patent Citations
Physical distribution planning system, physical distribution planning method, and physical distribution planning program
JP2009208933A
Matrix decomposition device and matrix decomposition method
JP2019079291A
Cited By
Multi-domain adaptive target tracking method and system based on domain customization adaptation
CN121074091A
A multi-domain adaptive target tracking method and system based on domain customization and adaptation
CN121074091B
