Clustering method, apparatus, and readable storage medium for federated graphs based on distributed graph embeddings

Abstract

The present invention provides a federated graph clustering method, device and readable storage medium based on distributed graph embedding. The method includes: constructing a first diagram based on the data of a first party, constructing a second diagram based on the data of a second party, encrypting the data of the first party and the data of the second party to obtain intersections to determine common nodes in the first diagram and the second diagram, and associating the first diagram and the second diagram based on the common nodes to obtain an federated graph; learning the federated graph using a distributed graph embedding algorithm based on random walks to determine an embedding vector [PiA,PiB] of the first diagram starting from the first diagram and an embedding vector [PiA',PiB'] of the second diagram starting from the second diagram; and performing clustering analysis of the embedding vector [PiA,PiB] of the first diagram and the embedding vector [PiA',PiB'] of the second diagram of the federated graph based on a federated clustering method to obtain a clustering result. The method can realize federated graph clustering of both privacy data and obtain a better clustering effect.

Description

[Technical Field]

[0001] This application claims priority to the Chinese patent application filed on 28 January 2022, application number 202210106101.1, with the title of the invention "Method, apparatus and readable storage medium for federated graph clustering based on distributed graph embedding," the disclosures of which are incorporated herein by reference.

[0002] This invention belongs to the field of clustering and specifically relates to a Federated Graph Clustering method, apparatus, and readable storage medium based on Distributed Graph Embedding. [Background technology]

[0003] This section is intended to provide background or context to the embodiments of the invention described in the claims. The descriptions herein are not considered prior art simply because they are included in this section.

[0004] Currently, Federated Learning (EML) has high application potential for data sharing when data is unavailable and for mining the value of multi-faceted data. However, it is mainly supported by traditional machine learning algorithms such as classification and regression models, which focus on individual value assessment and lack mining for potential group behavior. At the same time, since graph computation is related to the intercomputation of multi-ring topologies of multi-faceted data, research on the development of graph mining algorithms based on privacy computation is currently weak, and there are few industry achievements. Therefore, associative learning based on privacy graph structures is a problem that needs to be solved urgently. [Overview of the project] [Means for solving the problem]

[0005] To address the problems present in the aforementioned prior art, a federated graph clustering method, apparatus, and computer-readable storage medium based on distributed graph embedding have been proposed. This method, apparatus, and computer-readable storage medium can solve the aforementioned problems.

[0006] The present invention provides the following solutions.

[0007] In the first embodiment, a federated graph clustering method based on distributed graph embeddings is provided. This method involves constructing a first figure based on first-party data and a second figure based on second-party data, encrypting the first-party data and second-party data to find crossovers, determining common nodes in the first and second figures, associating the first and second figures based on the common nodes to obtain a federated graph, learning the federated graph using a distributed graph embedding algorithm based on a random walk, and determining the embedding vectors of the first figure [PiA, PiB] which are the embedding vectors of each node in the first figure starting from the first figure, and the second This method includes determining the embedding vectors [PiA',PiB'] of Figure 2, which are the embedding vectors of each node in Figure 2, starting from Figure 2, where PiA and PiA' are the embedding vectors of each node in Figure 1, and PiB and PiB' are the embedding vectors of each node in Figure 2; and performing clustering analysis on the embedding vectors [PiA,PiB] of Figure 1 and the embedding vectors [PiA',PiB'] of Figure 2 of the federated graph based on a federated clustering method to obtain the clustering results.

[0008] In one embodiment, determining the embedding vectors [PiA, PiB] in the first figure and the embedding vectors [PiA’, PiB’] in the second figure includes randomly walking multiple times in the association graph with the nodes in the first figure as the starting nodes, where the first person determines PiA based on the walk path in the first figure, the second person determines PiB based on the matching walk path in the second figure, randomly walking multiple times in the association graph with the nodes in the second figure as the starting nodes, where the second person determines PiB’ based on the walk path in the second figure, and the first person determines PiA’ based on the matching walk path in the first figure.

[0009] In one embodiment, obtaining an association graph by associating the first and second figures based on common nodes further includes removing isolated nodes that have no direct or indirect relationship with the common nodes in the first and second figures to obtain the association graph.

[0010] In one embodiment, the data of the first person and the data of the second person are isolated from each other.

[0011] In one embodiment, the nodes in the first figure are the first person's user and / or the first person's merchant, and the edges in the first figure are determined based on the relationship between the nodes in the first figure. The nodes in the second figure are the second person's user and / or the second person's merchant, and the edges in the second figure are determined based on the relationship between the nodes in the second figure.

[0012] In one embodiment, encrypting the data of the first person and the data of the second person to obtain an intersection and determining the common nodes in the first figure network and the second figure network includes aligning the common nodes in the first figure network and the second figure network based on the attribute information of the merchant and / or the user.

[0013] In one embodiment, the first player randomly walks multiple times in the federated graph, starting from a node in Figure 1 as the starting node. The first player determines PiA based on the walk path in Figure 1, defining the number of random walk steps M. The first player's data randomly walks in Figure 1, starting from any node in Figure 1 as the starting node. When the first player walks to any of the common nodes, the walk stops, and the number of walk steps in Figure 1, Mia, and the identifier of the common node walked to, Vab, are determined. i This includes recording each node of Figure 1 that was walked through, performing a random X walk, statistically calculating the number of walk steps Mia of Figure 1 and the frequency of each node in Figure 1 being walked each time, obtaining a node frequency matrix of Figure 1 corresponding to the number of walk steps Mia of Figure 1, and performing a matrix cumulative calculation of the node frequency matrix of Figure 1 corresponding to the number of walk steps Mia of Figure 1, dividing by the number of random walks X, to obtain the Figure 1 portion PiA of the embedding vector of Figure 1.

[0014] In one embodiment, the second party determines PiB based on the matching walk path in the second figure, which is the identifier Vab of each common node walked by the first party during or after X random walks. i The second party sends the number of walk steps Mia for all corresponding first diagrams, and the second party sends each common node Vab i Starting from the first point, the graph embedding vector PiB_Vab walks through the corresponding (M-Mia) steps in the second figure. i Confirmation of all common nodes Vab i The corresponding graph embedding vector PiB_Vab i This involves accumulating the values ​​and dividing by the number of subwalks X1, which is the number of times the second figure was reached in X random walks, to obtain the second figure portion PiB of the embedding vector for the first figure.

[0015] In one embodiment, the node frequency matrix of the first figure corresponding to the number of walk steps Mia in each first figure is PA_Mia=[ Pa n1including "_Mia, n1 = 1, 2,..., Na", where the first figure includes Na nodes Pa n1 where Mia is an integer between the minimum number of steps m and the total number of steps M from the starting node to the common node, and Pa n _After Mia randomly takes X steps from the starting node, it is the number of times passing through the node Pa in the first figure n of the first figure

[0016] In one embodiment, PiA is calculated by the following formula:

[0017] [Number]

[0018] In one embodiment, PiB is obtained by calculating with the following formula:

[0019] [Number]

[0020] In one embodiment, taking the nodes of the second figure as the starting nodes, randomly walking multiple times in the associated graph, the second person determines PIB' based on the walk path in the second figure. Define the random walk step number M'. When the second person randomly walks in the second figure with any node of the second figure as the starting node and walks to any common node, stop the walk. The walk step number Mia' of the second figure, the identifier Vab of the common node walked through iThis includes recording each node of the second figure that was walked this time, randomly walking X' times, statistically calculating the number of walk steps Mia' of the second figure walked each time and the frequency of each node of the second figure being walked to obtain a node frequency matrix of the second figure corresponding to the number of walk steps Mia' of the second figure, and performing matrix cumulative calculation on the node frequency matrix of the second figure corresponding to the number of walk steps Mia' of the second figure, dividing by the number of random walks X' to obtain the second figure portion PIB' of the embedding vector of the second figure.

[0021] In one embodiment, the first party determines PiA' based on the matching walk path in Figure 1, either during or after X' random walks, and the second party determines the identifier Vab of each common node walked. i The first party then sends the number of walk steps Mia' for all corresponding second diagrams, and the first party sends each common node Vab i Starting from the first figure, the graph embedding vector PiA'_Vab walks through the corresponding (M'-Mia') steps. i Confirmation of all common nodes Vab i The corresponding graph embedding vector PiA'_Vab i This includes accumulating the results and dividing by the number of subwalks X'1 to obtain the first figure portion PiA' of the embedding vector of the second figure. Here, the number of subwalks X'1 is the number of times the first figure is reached during the process of X' random walks.

[0022] In one embodiment, the node frequency matrix in the second figure corresponding to each walk step number Mia' is:

number

[0023] In one embodiment, PIB' is calculated by the following formula:

[0024]

number

[0025] In one embodiment, PiA' is obtained by the following formula:

[0026]

number

[0027] In one embodiment, performing clustering analysis on the embedding vectors [PiA, PiB] of the first figure and [PiA', PiB'] of the second figure of a federated graph based on a federated clustering method includes: performing clustering analysis on the first figure portion PiA of the embedding vector of the first figure and the first figure portion PiA' of the embedding vector of the second figure to obtain a first cluster of the first figure portion of the federated graph; performing clustering analysis on the second figure portion PiB of the embedding vector of the first figure and the second figure portion PiB' of the embedding vector of the second figure to obtain a second cluster of the first figure portion of the federated graph; and filtering multi-figure clusters based on the first and second clusters to obtain a target cluster with a higher screening level.

[0028] In the second embodiment, there is a configuration module that constructs Figure 1 based on the first party's data and Figure 2 based on the second party's data, an association module that encrypts the first party's data and the second party's data to find crossovers, determines common nodes in Figure 1 and Figure 2, associates Figure 1 and Figure 2 based on the common nodes to obtain a federated graph, and learns the federated graph using a distributed graph embedding algorithm based on a random walk, and the embedding vector [PiA, PiB] of Figure 1 starting from Figure 1 and the embedding vector of Figure 2 starting from Figure 2 A federated graph embedding-based federated graph clustering device is provided, which includes a learning module for determining the vectors [PiA',PiB'] (where PiA and PiA' are the embedding vectors for each node in Figure 1, and PiB and PiB' are the embedding vectors for each node in Figure 2), and a clustering module for performing clustering analysis of the embedding vectors [PiA,PiB] in Figure 1 and the embedding vectors [PiA',PiB'] in Figure 2 of the federated graph based on a federated clustering method to obtain clustering results.

[0029] In a third embodiment, a federated graph clustering device based on distributed graph embedding is provided, which includes at least one processor and a memory communicated with the at least one processor, wherein the memory stores instructions executable by at least one processor (such as the method of the first embodiment) so that the at least one processor can execute them.

[0030] In a fourth embodiment, a computer-readable storage medium is provided which, when executed by a multicore processor, stores a program that causes the multicore processor to perform the method as in the first embodiment.

[0031] One of the advantages of the above embodiment is that by utilizing a distributed graph embedding algorithm, it is possible to learn graph embedding vectors of a federated graph under the condition that the first party's data and the second party's data are mutually private data, the graph structure topology properties of the federated graph can be reduced to a matrix, the complexity of the computation can be reduced by matrix analysis, and the effectiveness and efficiency of federated graph computation can be increased.

[0032] Other advantages of the present invention will be described in more detail in accordance with the following description and drawings.

[0033] It should be understood that the above description is merely an outline of the technical means of the present invention, so that they can be more clearly understood and thus implemented in accordance with the specification. In order to make the aforementioned and other objectives, features and advantages of the present invention easier to understand, specific embodiments of the present invention will be described below with specific examples. Those skilled in the art will understand the advantages and merits of this specification, as well as other advantages and merits, by reading the detailed description of the exemplary embodiments below. The drawings are used solely to illustrate exemplary embodiments and are not intended to limit the invention. Furthermore, the same parts are indicated by the same reference numerals throughout the drawings.

[0034] In drawings, identical or corresponding reference numerals indicate identical or corresponding parts. [Brief explanation of the drawing]

[0035] [Figure 1] This is a schematic diagram of the configuration of a federated graph clustering device based on distributed graph embedding according to one embodiment of the present invention. [Figure 2] This is a schematic diagram of the flow of a federated graph clustering method based on distributed graph embedding according to one embodiment of the present invention. [Figure 3] These are schematic diagrams of the first and second figures according to one embodiment of the present invention. [Figure 4] This is a schematic diagram of an association graph according to one embodiment of the present invention. [Figure 5] This is a schematic diagram of another association graph according to one embodiment of the present invention. [Figure 6] This is a schematic diagram of the configuration of a federated graph clustering device based on distributed graph embedding according to one embodiment of the present invention. [Figure 7] This is a schematic diagram of the configuration of a federated graph clustering device based on distributed graph embedding according to another embodiment of the present invention. [Modes for carrying out the invention]

[0036] The exemplary embodiments of this disclosure will be described in more detail below with reference to the accompanying drawings. While the drawings show exemplary embodiments of this disclosure, it should be understood that this disclosure should not be limited to the embodiments described herein and can be realized in various forms. On the contrary, these embodiments are provided to allow for a more complete understanding of this disclosure and to fully convey the scope of this disclosure to those skilled in the art.

[0037] In the description of the embodiments of this application, terms such as “includes” or “having” are intended to indicate the presence of features, figures, steps, actions, parts, sections, or combinations thereof disclosed herein, and not to exclude the possibility of the presence of one or more other features, figures, steps, actions, parts, sections, or combinations thereof.

[0038] Unless otherwise specified, " / " means OR, for example A / B can represent A or B, and "and / or" in this specification is merely a relational relationship that describes the related objects, and can indicate that there may be three relationships, for example A and / or B, that A exists alone, that A and B exist together, and that B exists alone.

[0039] Terms such as "first," "second," etc., are used solely to describe the purpose and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features described. Therefore, features limited to "first," "second," etc., may explicitly or implicitly include one or more features. In the description of the embodiments of this application, "multiple" means two or more unless otherwise specified.

[0040] All code in this application is illustrative, and those skilled in the art can conceive of various variations without deviating from the spirit of this application, based on factors such as the programming language used, specific needs, and personal habits.

[0041] In cases where there is no conflict, the embodiments and features of the present invention may be combined with each other. The present invention will be described in detail below with reference to the drawings, combining embodiments.

[0042] As shown in Figure 1, Figure 1 is a configuration diagram of the hardware operating environment according to the solution of an embodiment of the present invention.

[0043] Figure 1 should be described as a configuration diagram of the hardware operating environment of a federated graph clustering device based on distributed graph embedding. The federated graph clustering device incorporating the embodiment of the present invention based on a distributed graph may be a PC or a terminal device such as a mobile computer.

[0044] As shown in Figure 1, the federated graph clustering device based on distributed graph embedding includes a processor 1001 such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002. Here, the communication bus 1002 is used to enable connectivity communication between these components. The user interface 1003 includes input units such as a display and a keyboard, and optionally the user interface 1003 includes a standard wired interface and a wireless interface. The network interface 1004 optionally includes a standard wired interface and a wireless interface (e.g., a Wi-Fi interface). The memory 1005 may be high-speed RAM memory or stable memory (non-volatile memory), such as disk memory. The memory 1005 may optionally be a storage device independent of the processor 1001.

[0045] Those skilled in the art will understand that the structure of the federated graph clustering device based on distributed graph embedding shown in Figure 1 is not an limitation to federated graph clustering devices based on distributed graph embedding, and that it may include more or fewer components than shown, or a combination of some components, or a different arrangement of components.

[0046] As shown in Figure 1, the memory 1005, which is a computer storage medium, may include an operating system, a network communication module, a user interface module, and a distributed graph embedding federated graph clustering program. Here, the operating system is a program that manages and controls the hardware and software resources of the distributed graph embedding federated graph clustering device and supports the execution of the distributed graph embedding federated graph clustering program and other software or programs.

[0047] In the federated graph clustering device incorporated based on the distributed graph shown in Figure 1, the user interface 1003 is mainly used to receive requests and data sent from the first terminal, the second terminal, and the management terminal; the network interface 1004 is mainly used to connect to the background server and communicate with the background server; and the processor 1001 can be used to call the federated graph clustering program based on the distributed graph embedding stored in memory 1005 and to perform the following operations.

[0048] Based on the first party's data, Figure 1 is constructed, and based on the second party's data, Figure 2 is constructed. The first party's data and the second party's data are encrypted and cross-referenced to determine common nodes in Figures 1 and 2. Based on the common nodes, Figures 1 and 2 are associated to obtain a federated graph. The federated graph is trained using a distributed graph embedding algorithm based on random walks to determine the embedding vectors [PiA, PiB] of Figure 1, starting from Figure 1, and the embedding vectors [PiA', PiB'] of Figure 2, starting from Figure 2, where PiA and PiA' are the embedding vectors for each node in Figure 1, and PiB and PiB' are the embedding vectors for each node in Figure 2. Clustering analysis is performed on the embedding vectors [PiA, PiB] of Figure 1 and [PiA', PiB'] of Figure 2 of the federated graph based on a federated clustering method to obtain clustering results.

[0049] This allows us to learn graph embedding vectors for a federated graph under the condition that first-party and second-party data are mutually private data, by utilizing a distributed graph embedding algorithm. This enables us to reduce the topological properties of the federated graph's graph structure to a matrix, and then reduce the computational complexity through matrix analysis, thereby improving the effectiveness and efficiency of federated graph computations.

[0050] Figure 2 is a schematic flowchart of a federated graph computation method based on distributed graph embedding according to one embodiment of the present invention. In this flow, from the perspective of the device, the main execution entity may be one or more electronic devices, and more specifically, a processing module. From the perspective of the program, the main execution entity may be a program installed on these electronic devices. In this embodiment, the main execution entity of the method may be the processor in the embodiment shown in Figure 1.

[0051] As shown in Figure 2, the method provided by this embodiment includes the steps of 202a, constructing a first figure based on first-party data, and 202b, constructing a second figure based on second-party data.

[0052] The data of the first party and the data of the second party are, respectively, the first party's private data and the second party's private data. Furthermore, it can be understood that the first party constructs the first figure based on the first party's data, and the second party constructs the second figure based on the second party's data.

[0053] Referring to Figure 3, we can define the relationship between the entity nodes and edges in the diagram based on the actual task, and construct the first Figure A and the second Figure B, respectively, with the data owned by both.

[0054] In some embodiments, the nodes in the first diagram are first-party users and / or first-party merchants, and the edges in the first diagram are determined based on the relationships between the nodes in the first diagram; the nodes in the second diagram are second-party users and / or second-party merchants, and the edges in the second diagram are determined based on the relationships between the nodes in the second diagram. For example, the relationships may be transaction relationships between users, transfer relationships between users, or any other relationships between nodes, such as transfer relationships between merchants.

[0055] 204. Encrypt the data of the first party and the data of the second party and seek cross-relationships to determine common nodes in Figure 1 and Figure 2. Based on the common nodes, Figure 1 and Figure 2 are associated to obtain an association graph.

[0056] By finding common nodes in Figure 1 and Figure 2, we can establish a relationship between Figure 1 and Figure 2, where common nodes are the same entity nodes that are common to both Figure 1 and Figure 2, such as the same user or the same merchant.

[0057] Referring to Figure 4, V in Figure A of the first figure AB1 and V in the second figure AB1 This is a common node, and V in Figure 1A. AB2 and V in the second figure AB2 We can assume that these are common nodes, thereby relating Figure A (first) and Figure B (second) to form a federated graph.

[0058] In some embodiments, common nodes in the first and second diagram networks may be aligned based on merchant and / or user attribute information. For example, common nodes corresponding to the same user can be determined from user attribute information such as mobile phone number or email address.

[0059] In some embodiments, after relating the first and second figures, it is also possible to obtain a federated graph by removing isolated nodes that have no direct or indirect relationship with the common nodes in the first and second figures.

[0060] Referring to Figure 4, since the aforementioned isolated island nodes cannot perform federated training spanning Figure 1 and Figure 2, a federated graph for computing both figures can be formed by filtering out the isolated island nodes in Figure 1A and Figure 2B.

[0061] 206. The federated graph is trained using a distributed graph embedding algorithm based on random walks, and the embedding vector [PiA, PiB] of Figure 1, starting from Figure 1, and the embedding vector [PiA', PiB'] of Figure 2, starting from Figure 2, are determined.

[0062] PiA and PiA' are the embedding vectors for each node in Figure 1, and PiB and PiB' are the embedding vectors for each node in Figure 2.

[0063] Graph embedding is a method of mapping graph nodes (high-dimensional vectors) to low-dimensional vectors, thereby obtaining a representation of the uniqueness of each node and enabling tasks such as recommendation, classification, and prediction using these vectors. A graph embedding algorithm based on a random walk first samples nodes in the graph multiple times using a random walk algorithm to obtain several node sequences, and then generates a vector representation of each node in the graph based on these node sequences to obtain a graph embedding vector. Specifically, in this embodiment, the embedding vector [PiA, PiB] of the first graph includes the first graph portion PiA and the second graph portion PiB. The first graph portion PiA is obtained by sampling in the first graph by a first party, and the second graph portion PiB is obtained by sampling in the second graph by a second party. Therefore, in order to obtain the first and second graph portions respectively, it is necessary to distribute the graph embedding method to the first and second graphs using a distributed graph embedding algorithm.

[0064] First, we define the number of random walk steps for each node in the graph embedding as M steps. For each node in the first figure, if we randomly walk M steps, there is a possibility of walking from the first figure through a common node to the second figure, where we walk Mia steps in the first figure A and (M-Mia) steps in the second figure B. For example, referring to Figure 5, assuming we start from node 1, we walk M=4 steps, passing through nodes 1-2-5-6-7, walking 2 steps in the first figure A, i.e., 1-2-5, and walking 2 steps in the second figure B, i.e., 5-6-7.

[0065] In some embodiments, the 204 further includes randomly walking multiple times in the federated graph with a node in the first figure as the starting node, the first person determining PiA based on the walk path in the first figure, the second person determining PiB based on the matching walk path in the second figure, randomly walking multiple times in the federated graph with a node in the second figure as the starting node, the second person determining PiB' based on the walk path in the second figure, and the first person determining PiA' based on the matching walk path in the first figure.

[0066] The matching walk path in the second figure above is obtained by starting from a common node between the first and second figures and walking randomly multiple times in the second figure.

[0067] The matching walk path in the first figure is obtained by starting from a common node between the first and second figures and walking randomly multiple times in the first figure.

[0068] In a federated graph, the random walk path in the second figure is unknown to the first player, and similarly, the random walk path in the first figure is unknown to the second player. Therefore, when walking randomly starting from a node in the first figure, the first player can obtain the PiA portion of the embedding vector in the first figure, and the second player can obtain the PiB portion of the same embedding vector by matching it with the remaining walk task sent from the first player. Similarly, when walking randomly starting from a node in the second figure, the second player can obtain the PiB' portion of the embedding vector in the second figure, and the first player can obtain the PiA' portion of the embedding vector in the second figure by matching it with the remaining walk task sent from the second player.

[0069] In some examples, determining the first figure portion PiA of the embedding vector of the first figure is A random number of walk steps M is defined randomly. The first player randomly walks through Figure 1, starting from any of the nodes in Figure 1. When the player reaches any of the common nodes, the walk stops, and the number of walk steps Mia in Figure 1 and the identifier Vab of the common node walked to are recorded. i This includes recording each node in Figure 1 that was walked through this time.

[0070] After randomly walking X times, the number of walk steps Mia in Figure 1 and the frequency of each node in Figure 1 being walked are statistically calculated, and a node frequency matrix of Figure 1 corresponding to the number of walk steps Mia in Figure 1 is obtained.

[0071] The node frequency matrix in Figure 1 corresponding to the number of walk steps Mia in each Figure 1 is calculated using matrix accumulation, and then divided by the number of random walks X to obtain the Figure 1 portion PiA of the embedding vector in Figure 1.

[0072] Specifically, obtaining the node frequency matrix in Figure 1 corresponding to the number of walk steps Mia in each Figure 1 means that Once you walk and reach a common node, Mia=1, the frequency matrix is ​​PA_1=[Pa1_1,Pa2_1,Pa3_1,Pa4_1,…,Pa Na _1] and Reaching the common node by walking twice, Mia=2, the frequency matrix is ​​PA_2=[Pa2_2,Pa2_2,Pa3_2,Pa4_2,…,Pa Na _2] is, ... The objective is to reach a common node by taking M walks. Mia=M, the frequency matrix is ​​PA_M=[ Pa M _M,Pa2_M,Pa3_M,Pa4_M,…,Pa Na This includes the fact that it is [M].

[0073] As described above, starting from Figure 1, we have the common node Vab i The frequency matrix PA_Mia is obtained by walking through Mia steps up to this point. Here, Mia is an integer between the minimum number of steps m from the starting node to the common node and the total number of steps M.

number

[0074] Here, Figure 1 shows Na nodes Pa n1 It includes, where Mia is an integer between the minimum number of steps m from the starting node to the common node and the total number of steps M, and Pa n After _Mia performs Mia steps randomly X times from the starting node, node Pa in Figure 1 n This is the number of times it passed through.

[0075] Furthermore, the first figure portion PiA of the first embedding vector can be calculated by the following formula:

[0076]

number

[0077] In some embodiments, determining the second figure portion PiB of the embedding vector of the first figure is: During or after X random walks, the first party identifies each common node walked through by Vab. i Then, the number of walk steps Mia for all corresponding Figure 1 is sent to the second party. For example, the second party receives multiple (Vab) from the first party. i This includes receiving the remaining combinations of walk step numbers (Mib = M - Mia) shown in Figure 2.

[0078] Furthermore, the second party will select each common node Vab i Starting from the first, the graph embedding vector PiB_Vab is obtained by walking the corresponding Mib=(M‐Mia) steps in the second. i Confirm.

[0079] For example, in Figure 1A, the common node Vab i After walking through the Mia steps, the common node Vab is shown in Figure 2. i Starting from the walk point, we continue walking through the Mib steps, and the frequency matrix is

number

[0080] The common node Vab mentioned above i The corresponding graph embedding vector PiB_Vab i It is also understandable that this can be calculated in advance from the second figure.

[0081] Furthermore, all common nodes Vab i The corresponding graph embedding vector PiB_Vab i The values ​​are accumulated and divided by the number of subwalks X1 to obtain the second figure portion PiB of the embedding vector of the first figure.

[0082] For example, PiB is

number

[0083] Here, the subwalk count X1 is the number of times the path has reached Figure 2 during X random walks.

[0084] Similarly, by randomly walking multiple times in the federated graph with the node in the second figure as the starting node, a second party can determine PiB' based on the walk path in the second figure, and a first party can determine PiA' from the matching walk path in the first figure, thereby obtaining the embedding vector [PiA',PiB'] in the second figure.

[0085] Specifically, we can define the number of random walk steps M'. The second player starts at any of the nodes in Figure 2 and performs a random walk in Figure 2. When they reach any of the common nodes, they stop walking and record the number of walk steps Mia' in Figure 2 and the identifier Vab of the common node they walked to. i Then, the nodes of Figure 2 that were walked this time are recorded, and after walking X' times randomly, the number of walk steps Mia' in Figure 2 and the frequency of each node in Figure 2 being walked are statistically calculated to obtain the Figure 2 node frequency matrix corresponding to the number of walk steps Mia' of each Figure 2, the Figure 2 node frequency matrix corresponding to the number of walk steps Mia' of each Figure 2 is cumulatively calculated and divided by the number of random walks X' to obtain the Figure 2 portion PIB' of the Figure 2 embedding vector.

[0086] In some embodiments, the first party determines PiA' based on the matching walk path in Figure 1, During or after X' random walks, the second party identifies each common node walked through by Vab. iAnd, the number of walk steps Mia' in all corresponding Figure 2 is sent to the first party, The first party is each common node Vab i Starting from there, the graph embedding vector PiA'_Vab was obtained by walking through the corresponding (M'-Mia') steps in Figure 1. i To confirm and All common node Vab i The corresponding graph embedding vector PiA'_Vab i This includes accumulating the values ​​and dividing by the number of subwalks X'1 to obtain the first figure portion PiA' of the embedding vector of the second figure.

[0087] Here, the subwalk count X'1 refers to the number of times the path has reached Figure 1 during the X' random walk process.

[0088] In some embodiments, the node frequency matrix in the second figure corresponding to each walk step number Mia' is:

number

[0089] Here, the second figure shows Nb nodes Pb n2 The second diagram shows that Mia' takes an integer between the minimum number of steps m' from the starting node to the common node and the total number of steps M', and Pb2_Mia' randomly walks X' Mia' steps from the starting node, and then the node Pb n2 This is the number of times it passed through.

[0090] In some embodiments, PIB' is calculated using the following formula:

[0091]

number

[0092] In some examples, PIA' can be calculated using the following formula:

[0093]

number

[0094] 208. Based on the federated clustering method, the embedding vectors [PiA, PiB] in Figure 1 and [PiA', PiB'] in Figure 2 of the federated graph are subjected to clustering analysis to obtain the clustering results.

[0095] In some embodiments, the 205 further includes: performing clustering analysis on the Figure 1 portion PiA of the embedding vector of Figure 1 and the Figure 1 portion PiA' of the embedding vector of Figure 2 based on a federated clustering method to obtain a first cluster of the Figure 1 portion of the federated graph; performing clustering analysis on the Figure 2 portion PiB of the embedding vector of Figure 1 and the Figure 2 portion PiB' of the embedding vector of Figure 2 based on a federated clustering method to obtain a second cluster of the Figure 1 portion of the federated graph; and screening multi-figure clusters based on the first and second clusters to obtain a target cluster with a higher cluster level.

[0096] In one example, the embodiment of the present invention can be used to identify risk groups spanning scenarios such as transaction fraud between payment institutions. Taking the example of mining fraudulent users of a certain payment application on payment platform A, there is currently a risk of a fraudulent group where a merchant on one payment channel trades profits, launders the money, and then transfers the transaction under the other payment channel. Transaction fraud can be associated by the following steps:

[0097] 1. Construct the first figure based on the first party's data, and construct the second figure based on the second party's data.

[0098] For example, a transfer transaction between payment platform A, a merchant, and a user forms the first diagram, and a transfer transaction between a user of payment platform B, a merchant, and another user forms the second diagram, where the entities are user and merchant nodes, and the relationships are transfer relationships between users and transaction-payment relationships between users and merchants.

[0099] 2. Encrypt the data of the first party and the data of the second party, seek cross-relationships, determine common nodes in Figure 1 and Figure 2, associate Figure 1 and Figure 2 based on the common nodes, and obtain an association graph.

[0100] For example, by using user identity information and company business information as primary keys, common users and merchants across two payment channels are aligned. Common users are the same registered transaction users across both payment channels, common merchants are aggregated payment merchants, and transaction merchants enable mutual recognition and scanning of merchant deposit codes.

[0101] 3. Train the association graph using a distributed graph embedding algorithm based on random walks.

[0102] This defines that each user or merchant node of payment platform B walks through step M, and after walking through step Mia, the shared node Vab i The frequency vector Pa-Mia, which reaches the shared node after passing through multiple Mia steps, is recorded. The embedding vector PiA is obtained by weighting and aggregating the vectors that reach the shared node through different steps. Based on the sequence of shared nodes reached, the vector matrix of the user of payment platform A walking through M-Mia steps from the shared node is matched in relation to the corresponding shared node Vab at the intermediate node. iThen, the vector matrix obtained by walking through (M-Ma) steps is encrypted and matched, and the results of multiple matching attempts are weighted and aggregated to obtain the corresponding embedding vector PiB, and PiA and PiB are stored in their respective data spaces.

[0103] Using a similar method, we randomly walk from a user node or merchant node on payment platform A to obtain PiA' and PiB'.

[0104] 4. Adopting a method based on associative learning clustering analysis, first, clustering analysis is performed on parts PiA and PiA' in Figure 1, respectively, and clusters mined based on the vectors in Figure 1 are marked. Furthermore, intermediate nodes record data about the central nuclei of the clusters. After introducing parts PiB and PiB' in Figure 2, each node within the cluster compares the cluster situation in Figure 2, mines the central nuclei of the clusters in the vector space of Figure 2, filters the clusters again, and obtains the most related node data.

[0105] In this specification, any reference to terms such as “several possible embodiments,” “several examples,” “examples,” “specific examples,” or “several examples” means that the specific features, structures, materials, or characteristics described in relation to that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the general expressions of the above terms do not necessarily apply to the same embodiment or example. Furthermore, any specific features, structures, materials, or characteristics described may be combined in an appropriate manner in any one or more embodiments or examples. Furthermore, where inconsistent, a person skilled in the art may combine or combine different embodiments or examples and features of different embodiments or examples described herein.

[0106] Furthermore, terms such as "first," "second," etc., are used solely to describe the purpose and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features shown. Therefore, features limited to "first," "second," etc., may explicitly or implicitly include one or more features. In the description of the embodiments of this application, "multiple" means two or more unless otherwise specified.

[0107] The descriptions of processes or methods in the flowchart or otherwise described herein represent modules, segments, or portions representing one or more codes of executable instructions that include steps for realizing a particular logical function or process, and the scope of preferred embodiments of the present invention does not necessarily follow the illustrated or discussed order, including performing functions essentially concurrently or in reverse order based on the relevant functions. This should be understood by those skilled in the art of embodiments of the present invention.

[0108] In the method flowcharts of embodiments of the present application, some operations are described as different steps performed in a certain order. Such flowcharts are illustrative and not limiting. Some of the steps described herein may be grouped together, performed as a single operation, some steps may be divided into multiple substeps, and may be performed in an order different from that shown herein. The various steps shown in the flowcharts can be implemented in any way by any circuit structure and / or tangible mechanism (e.g., software, hardware (e.g., logic functions implemented by a processor or chip) and / or any combination thereof) that operates in a computer device.

[0109] Based on the same technical concept, an embodiment of the present invention provides a federated graph clustering device based on distributed graph embeddings for performing a federated graph clustering method based on distributed graph embeddings provided in any of the embodiments described above. Figure 6 is a schematic diagram of the federated graph clustering device based on distributed graph embeddings provided in an embodiment of the present invention.

[0110] As shown in Figure 6, the apparatus 600 is A configuration module 601 for constructing a first figure based on the first party's data and a second figure based on the second party's data, The association module 602 encrypts the data of the first party and the data of the second party to seek cross-relationships, determines common nodes in the first and second figures, and associates the first and second figures based on the common nodes to obtain a federated graph. A learning module 603 is used to learn a federated graph using a distributed graph embedding algorithm based on random walks, and to determine the embedding vectors [PiA, PiB] for Figure 1, starting from Figure 1, and the embedding vectors [PiA', PiB'] for Figure 2, starting from Figure 2 (where PiA and PiA' are the embedding vectors for each node in Figure 1, and PiB and PiB' are the embedding vectors for each node in Figure 2), The system includes a clustering module 604 for performing clustering analysis on the embedding vectors [PiA, PiB] of the first figure and [PiA', PiB'] of the second figure of the federated graph based on a federated clustering method to obtain clustering results.

[0111] It is necessary to explain that the apparatus in the embodiment of this application can implement various processes of the embodiment of the method described above and achieve similar effects and functions. Further explanation is omitted here.

[0112] Figure 7 shows a federated graph clustering device based on distributed graph embedding according to one embodiment of the present invention for performing the federated graph clustering method based on distributed graph embedding shown in Figure 2. The device includes at least one processor and a memory communicated with the at least one processor, the memory storing instructions executable by the at least one processor so that the at least one processor can perform the method of the embodiment described above.

[0113] According to some embodiments of the present application, a non-volatile computer storage medium is provided that stores computer executable instructions, when executed by a processor, that perform the method described in the above embodiments.

[0114] The various embodiments described herein are described cumulatively. Similar parts between each embodiment should be referenced to one another, and each embodiment focuses on its differences from the others. In particular, the embodiments of apparatus, devices, and computer-readable storage media are essentially similar to the embodiments of the method, and therefore their descriptions are simplified; relevant points should be referred to the partial descriptions in the embodiments of the method.

[0115] Since the apparatus, devices, and computer-readable storage media provided by the embodiments of this application correspond one-to-one with the methods, the apparatus, devices, and computer-readable storage media also have beneficial technical effects similar to the corresponding methods. As the beneficial technical effects of the methods have been described in detail above, no further explanation of the beneficial technical effects of the apparatus, devices, and computer-readable storage media will be given here.

[0116] Those skilled in the art will understand that embodiments of the present invention may be provided as methods, systems, or computer program products. Accordingly, the present invention may take the form of complete hardware embodiments, complete software embodiments, or embodiments combining software and hardware aspects. The present invention may also take the form of a computer program product implemented on one or more available storage media of a computer (including, but not limited to, disk memory, CD-ROM, optical memory, etc.) containing available program code for the computer.

[0117] The present invention will be described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in a flowchart and / or block diagram, as well as combinations of flows and / or blocks in a flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions are provided to a processor of a general-purpose computer, a dedicated computer, an embedded processor, or another programmable data processing device, and the instructions executed by the computer or other programmable data processing device processor can generate a machine which is a means for realizing one or more flows in a flowchart and / or one or more blocks in a block diagram.

[0118] These computer program instructions can also be stored in computer-readable memory that can operate a computer or other programmable data processing device in a particular way, thereby generating a product that includes an instruction unit that implements a function specified in one or more flows of a flowchart and / or one or more blocks of a block diagram.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device to perform a series of operational steps in the computer or other programmable device to generate processing to be performed by the computer, so that the instructions executed in the computer or other programmable device provide steps to perform a function specified in one or more flows of a flowchart and / or one or more blocks of a block diagram.

[0120] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0121] Memory can take various forms, including non-persistent memory in computer-readable media such as read-only memory (ROM) or flash memory (flash RAM), random-access memory (RAM), and / or non-volatile memory. Memory is an example of computer-readable media.

[0122] Computer-readable media include persistent, non-persistent, removable, and non-removable media, and information storage can be realized by any method or technique. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable, programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only optical disc memory (CD-ROM), digital multifunction optical disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape / disk storage devices or other magnetic storage devices or other non-transmission media that can be used to store information accessible by a computer. Furthermore, while the accompanying drawings describe the operation of the method of the present invention in a specific order, this does not require or imply that these operations must be performed in that specific order, or that the desired results cannot be achieved unless all operations are performed. Additionally or alternatively, several steps can be omitted, multiple steps can be merged into a single step execution, and / or a single step can be broken down into multiple step executions.

[0123] While the spirit and principles of the present invention have been described with reference to several specific embodiments, the present invention is not limited to the specific embodiments disclosed, and the classification of each embodiment does not mean that the features of these embodiments cannot be combined and benefited from; rather, the classification is merely for convenience of explanation. The present invention aims to encompass various amendments and equivalent configurations that fall within the spirit and scope of the appended claims. [Explanation of symbols]

[0124] 600 equipment 601 Configuration Modules 602 Related Modules 603 Learning Modules 604 Clustering Module 1001 Processor 1002 Communications Bus 1003 User Interface 1004 Network Interface 1005 memory

Claims

1. A federated graph clustering method performed by a processor, Construct a first figure, which is a graph showing the relationships between entity nodes and edges defined based on the first party's data, and construct a second figure, which is a graph showing the relationships between entity nodes and edges defined based on the second party's data. The data of the first party and the data of the second party are encrypted and their intersection is calculated, common nodes that are the same entity nodes present in both the first and second figures are identified, and the first and second figures are associated based on these common nodes to obtain a federated graph. The federated graph is learned using a distributed graph embedding algorithm based on a random walk, and the embedding vectors [PiA, PiB] of the first figure starting from the first figure and the embedding vectors [PiA', PiB'] of the second figure starting from the second figure are determined, where PiA and PiA' are the embedding vectors of the nodes in each of the first figures of the first figure, and PiB and PiB' are the embedding vectors of the nodes in each of the second figures of the second figure. This includes performing a federated graph clustering method based on the distributed graph embedding algorithm described above, thereby performing clustering analysis on the embedding vectors [PiA, PiB] of the first figure and the embedding vectors [PiA', PiB'] of the second figure of the federated graph, and obtaining clustering results. Encrypting the data of the first party and the data of the second party, and finding their intersection, and determining the common nodes that are the same entity nodes present in both the first and second figures, A federated graph clustering method based on distributed graph embedding, comprising aligning common nodes in the network of the first figure and the network of the second figure based on merchant and / or user attribute information.

2. Determining the embedding vectors [PiA, PiB] in the first figure and [PiA', PiB'] in the second figure is: The process involves performing multiple random walks in the federated graph, starting from the node in the first figure, determining PiA based on the walk path in the first figure, and determining PiB based on the matching walk path in the second figure. The method according to claim 1, comprising: performing a random walk multiple times in the federated graph with the node in the second figure as the starting node; the second person determining the PiB' based on the walk path in the second figure; and the first person determining the PiA' based on the matching walk path in the first figure.

3. Obtaining a federated graph by relating the first figure and the second figure based on the common node is, The method according to claim 1 or claim 2, further comprising removing isolated island nodes that have no direct or indirect relationship with the common nodes in the first and second figures to obtain the federated graph.

4. The method according to any one of claims 1 to 3, wherein the data of the first party and the data of the second party are isolated from each other.

5. The nodes in the first figure are first-party users and / or first-party merchants, and the edges in the first figure are determined based on the relationships between the nodes in the first figure. The method according to any one of claims 1 to 4, wherein the nodes in the second figure are second-party users and / or second-party merchants, and the edges in the second figure are determined based on the relationships between the nodes in the second figure.

6. The process involves performing a random walk multiple times in the federated graph, using the node in the first figure as the starting node, and determining PiA based on the first data's walk path in the first figure. The number of walk steps M for the random walk is defined, and the first person takes a random walk in the first figure, starting from any of the nodes in the first figure as the starting node. When the first person walks to any of the common nodes, the walk stops, and the number of walk steps M in the first figure is determined, and the identifier Vab of the common node walked to is determined. ｉ And, to record the nodes of each of the first diagrams that were walked through this time, After performing X random walks, the number of walk steps Mia of the first figure and the frequency of walks for each node in the first figure are statistically calculated to obtain a node frequency matrix for the first figure corresponding to each number of walk steps Mia of the first figure. The method according to any one of claims 2 to 5, comprising: performing a matrix accumulation calculation on the node frequency matrix of the first figure corresponding to the number of walk steps Mia of each of the first figures, and dividing by the number of walks X of the random walk to obtain the first figure portion PiA of the embedding vector of the first figure.

7. The fact that the aforementioned second party determines PiB based on the matching walk path in the second figure is that In or after the aforementioned X random walk process, the identifier Vab of each common node walked by the first party. ｉ And, the number of walk steps Mia for all corresponding first figures is sent to the second party, The aforementioned second party each of the aforementioned common node Vab ｉ Starting from there, the graph embedding vector PiB_Vab is obtained by walking the corresponding number of walk steps Mib in the second figure. ｉ We confirm that M = Mib + Mia, All of the aforementioned common node Vab ｉ The corresponding graph embedding vector PiB_Vab ｉ The number of steps taken is accumulated, and the subwalk count X is the number of times the walk reached the second figure in the X random walks. １ The method according to claim 6, comprising dividing by to obtain a second figure portion PiB of the embedding vector of the first figure.

8. The node frequency matrix of the first figure corresponding to the number of walk steps Mia in each of the first figures is PA_Mia = [Pa ｎ１ _Mia, n1 = 1, 2, ..., Na] Here, the first figure includes Na nodes Pa ｎ１ and The aforementioned Mia takes an integer between the minimum number of steps m from the starting node to the common node and the total number of steps M. The aforementioned Pa ｎ _Mia performs the Mia step randomly X times from the starting node, and then the node Pa in the first figure above ｎ The method according to claim 6 or claim 7, which is the number of times that the condition has been passed.

9. The aforementioned PiA is, [Math 1] The method according to claim 8, calculated by...

10. The aforementioned PiB is given by the formula [Math 2] The method according to claim 7, obtained by calculation.

11. The process involves performing multiple random walks in the federated graph, starting from the node in the second figure, and the second party determining PiB' based on the walk path in the second figure. The number of walk steps M' for a random walk is defined, and the second person takes a random walk in the second figure, starting from any of the nodes in the second figure as the starting node. When the second person walks to any of the common nodes, the walk stops, and the number of walk steps Mia' in the second figure and the identifier Vab of the common node that was walked are recorded. ｉ And, to record the nodes of each of the second diagrams that we walked through this time, After randomly walking X' times, the number of walk steps Mia' of the second figure and the frequency of walks for each node in the second figure are statistically calculated to obtain a node frequency matrix of the second figure corresponding to each walk step Mia' of the second figure. The method according to any one of claims 2 to 10, comprising: performing a matrix cumulative calculation on the node frequency matrix of the second figure corresponding to the number of walk steps Mia' of each of the second figures, and dividing by the number of walks X' of the random walk to obtain the second figure portion PiB' of the embedding vector of the second figure.

12. The first party determines PiA' based on the matching walk path in the first figure, During or after the X' random walk process, the second party identifies each common node walked through by Vab. ｉ And, send the number of walk steps Mia' for all corresponding second figures to the first party, The aforementioned first party each of the aforementioned common node Vab ｉ Starting from the first figure, the graph embedding vector PiA'_Vab is obtained by walking through the corresponding (M'-Mia') steps in the first figure. ｉ To confirm and All of the aforementioned common node Vab ｉ The corresponding graph embedding vector PiA'_Vab ｉ The number of steps is accumulated, and the number of subwalks X' is the number of times the walk reached the first figure during the process of the X' random walks. １ The method according to claim 11, comprising dividing by to obtain the first figure portion PiA' of the embedding vector of the second figure.

13. The node frequency matrix in the second figure corresponding to each of the walk step numbers Mia' is: [Math 3] Includes, Here, the second figure above shows Nb nodes Pb ｎ２ Includes, The Mia' takes an integer between the minimum number of steps m' from the common node to the starting node and the total number of steps M'. The aforementioned Pb ｎ２ _Mia' takes a random walk through X' Mia' steps from the starting node, and then reaches node Pb in the second figure. ｎ２ The method according to claim 11 or 12, which is the number of times it has passed through.

14. The above PiB' is given by the formula [Math 4] The method according to claim 13, calculated by...

15. The above PiA' is given by the formula [Math 5] The method according to claim 12, obtained by calculation.

16. Performing a federated graph clustering method based on the distributed graph embedding algorithm described above allows for clustering analysis of the embedding vectors [PiA, PiB] in the first figure and [PiA', PiB'] in the second figure of the federated graph, By performing a federated graph clustering method based on the distributed graph embedding algorithm, the first figure portion PiA of the embedding vector of the first figure and the first figure portion PiA' of the embedding vector of the second figure are subjected to clustering analysis to obtain the first figure portion PiA of the federated graph and the first cluster of the first figure portion PiA. By performing the federated graph clustering method based on the distributed graph embedding algorithm, the second figure portion PiB of the embedding vector of the first figure and the second figure portion PiB' of the embedding vector of the second figure are subjected to clustering analysis to obtain the second cluster of the second figure portion PiB and the second figure portion PiB' of the federated graph, The method according to any one of claims 1 to 15, further comprising filtering multi-drawing clusters based on the first cluster and the second cluster to obtain a target cluster with a higher clustering level.

17. A configuration module comprising: a first diagram which is a graph showing the relationships between entity nodes and edges defined based on the first party's data; and a second diagram which is a graph showing the relationships between entity nodes and edges defined based on the second party's data; An association module for obtaining a federated graph by encrypting the data of the first party and the data of the second party, finding the intersection, determining common nodes which are the same entity nodes that exist in both the first and second figures, and relating the first and second figures based on the common nodes, A learning module is used to learn the federated graph using a distributed graph embedding algorithm based on random walks, to determine the embedding vectors [PiA, PiB] of the first figure starting from the first figure and the embedding vectors [PiA', PiB'] of the second figure starting from the second figure, wherein PiA and PiA' are the embedding vectors of the nodes in each of the first figures of the first figure, and PiB and PiB' are the embedding vectors of the nodes in each of the second figures of the second figure, A clustering module is included for performing a clustering analysis of the embedding vectors [PiA, PiB] of the first figure and the embedding vectors [PiA', PiB'] of the second figure of the federated graph by executing a federated graph clustering method based on the distributed graph embedding algorithm, and obtaining clustering results. Encrypting the data of the first party and the data of the second party, and finding their intersection, and determining the common nodes that are the same entity nodes present in both the first and second figures, A federated graph clustering device based on distributed graph embedding, which includes aligning common nodes in the network of the first figure and the network of the second figure based on merchant and / or user attribute information.

18. At least one processor, A federated graph clustering device based on distributed graph embedding, comprising: a memory that is in communication with at least one processor and stores instructions executable by at least one processor so that at least one processor can perform the method according to any one of claims 1 to 16.

19. A computer-readable storage medium having stored a program that, when executed by a multicore processor, causes the multicore processor to execute the method according to any one of claims 1 to 16.

Images