32 results about "Cluster graph" patented technology

Provided is a process including: obtaining a clustered graph, the clustered graph having three or more clusters, each cluster having a plurality of nodes of the graph, the nodes being connected in pairs by one or more respective edges; determining visual attributes of cluster icons based on amounts of nodes in clusters corresponding to the respective cluster icons; determining positions of the cluster icons in a graphical visualization of the clustered graph; obtaining, for each cluster, a respective subset of nodes in the respective cluster; determining visual attributes of node icons based on attributes of corresponding nodes in the subsets of nodes, each node icon representing one of the nodes in the respective subset of nodes; determining positions of the node icons in the graphical visualization based on the positions of the corresponding cluster icons of clusters having the nodes corresponding to the respective node icons; and causing the graphical visualization to be displayed.

A method of detecting network worms include the following steps: (1) Profiling the TCP connection information collected from the protected network, quantifying the plurality of statuses contained in the TCP connection information; (2) Clustering the connection profiles to discover all the anomaly clusters that are specified by the condition composing of several adaptive thresholds; (3) Correlating the anomaly clusters to result in a new cluster graph or to extend an existing cluster graph; (4) Issuing a security incident about the worm propagation according to the propagation condition that also composes of several adaptive thresholds; and (5) Keeping and maintaining the status of the cluster graphs.

Provided is a process, including: obtaining a clustered graph, wherein each of the nodes has a plurality of respective node attributes other than an identifier of the node; obtaining a designation of a given node attribute from among the plurality of node attributes; identifying a first subset of nodes of the graph as having anomalous values of the given node attribute by comparing values of the given node attribute in the first subset to a distribution of the given node attribute; identifying a second subset of nodes of the graph as having representative values of the given node attribute by comparing values of the given node attribute in the second subset to the distribution of the given node attribute; and sending instructions to a client device to display a representation of the graph.

A method of compressing a graph representation, and method of solving a problem represented by a graph using graph compression enable improved data searching techniques. The graph representation includes a plurality of vertices and a plurality of edges. A first clustered graph is generated by clustering the plurality of vertices of the graph representation. Each cluster of vertices is then replaced by a clustered vertex, wherein the clustered vertex inherits edges from the vertices of the cluster. Superfluous edges of the first clustered graph are then removed. A traversal probability for each of the removed edges is determined, and one or more of the edges that have been removed is selected for embedding in the first clustered graph, based upon a traversal probability for each of the edges. Information relating to the one or more edges is then embedded into the first clustered graph.

The invention provides a layout cluster graph-based Android application repackage detection method. The method comprises the steps of: 1.1, dynamically executing an Android application and collecting user interface information including a layout in operation of the Android application and the change and jump of the layout occurring after the user triggers interaction behavior; 1.2, constructing a layout cluster graph according to the obtained user interface information contained in the application layout, wherein the layout cluster graph is used as a software birthmark of the application; 1.3, judging whether the application is subjected to repackage by comparing layout cluster graphs and obtaining the similarity. The software repackage detection method provides a strategy of dynamically executing Android applications, improves the resistance to two anti-detection methods of code obfuscation and encryption, and improves the accuracy of Android application repackage detection.

The invention discloses a network security anomaly detection algorithm based on a clustering graph neural network. The algorithm comprises the following steps: describing a network topology structureby using a graph model, optimizing node characteristics by using a graph neural network convolution layer, segmenting a graph into a plurality of disjoint sub-graphs by using a graph clustering algorithm, regarding each sub-graph as a node, regarding an adjacency relationship of the sub-graphs as an edge, forming a sub-graph, learning a weight for each node by utilizing a graph attention layer, performing weighted summation on features of all nodes in each sub-graph to form features of the nodes in the sub-graph, and finally judging whether a network is attacked or not by utilizing a full connection layer and a classifier layer. According to the method, a hierarchical graph neural network is constructed, node features in a graph are optimized through a graph convolution layer, local features on the graph are captured through a pooling layer based on a graph clustering algorithm, high-level semantic features are generated, situation features of the whole network are generated through afusion layer, and network situations are classified through a classifier.

The invention discloses a multi-satellite imaging task planning method, which comprises the following steps of: establishing a task model, and representing all point target tasks by using the task model; taking the orbital circles as a reference, taking tasks in the orbital circles as nodes of the clustering graph model, and constructing undirected edges among the nodes in the clustering graph model based on clustering constraint conditions to obtain an imaging task clustering graph model; aggregating the point target tasks meeting the clustering constraint condition into clustering tasks based on a heuristic rule, and calculating side swing angles of the clustering tasks based on a median theorem; constructing and utilizing constraint conditions and an objective function of task planningto construct a task planning directed acyclic graph model corresponding to the imaging task clustering graph model; and based on the task planning directed acyclic graph model, performing task planning by adopting a maximum and minimum ant colony algorithm to obtain a multi-satellite imaging task planning scheme. Under different data scales, a satisfactory task planning result can be obtained, andthe method has good stability.

The invention discloses a similarity data clustering method for dam safety monitoring data. The method comprises the following steps of separating a single measuring point sequence trend term from thehigh-frequency noise by utilizing an EMD algorithm, detecting the time sequence change points by adopting an inflection point detection method of a cumulative sum control graph, and splitting to obtain all subsequence sets; adopting a DTW distance measurement method for calculating the distance problem of the subsequence, and calculating the distance minimum value between the two pieces of subsequence data dynamically; and clustering the mined sub-time sequences by using hierarchical clustering, and dynamically analyzing the time sequence clustering distribution condition under different clustering numbers through the obtained tree-shaped clustering graph. According to the method, the similarity of the monitoring data is reasonably analyzed, the correlation of the same monitoring point inthe time sequence can be mined, and meanwhile the correlation between the safety monitoring data can be quantified. And the monitoring data subjected to similarity analysis processing can accuratelyreflect the change trend of the dam in the time dimension, and the subsequent monitoring data mining difficulty can be effectively reduced in combination with the change trend rule.

The invention discloses a subgraph searching method in a cluster graph set. The method comprises the steps: building an index, i.e., according to structural information and node attributes of cluster graphs, calculating similarity among all the cluster graphs in the cluster graph set, and clustering similar cluster graphs by a layer clustering algorithm according to the similarity among all the cluster graphs till one cluster graph is left; searching a subgraph, i.e., according to the structure and vertex attributes of a query graph started by a user, performing query graph isostructuralism matching on a cluster graph index tree from the top to the bottom of the tree. The tree-shaped index is built in a data set, data items not containing the query graph are filtered as early as possible, the query speed is accelerated, and the query efficiency is improved.

The invention discloses a link cluster graph based semantic position privacy protection method. The method comprises the following steps that 1) a users sends a present position, user privacy preference and the type of a request interest point to an anonymous server; 2) the anonymous server executes an anonymous algorithm to obtain an anonymous set; 3) the anonymous server sends an anonymous request to a position server; 4) the position server carries out query, and returns a query result to the anonymous server; and 5) the anonymous server carries out refinement on the query result, and returns a refinement result to a user. The method has the advantages that the success rate of anonymous request protection of the anonymous server is improved, anonymous time is reduced, and the query efficiency of the position server is improved.

The invention discloses a complex process module division method for large-scale customization. The complex process module division method includes the steps: based on the characteristic relationshipbetween the processes corresponding to a workpiece, providing a method for dividing a model on the process level; constructing a fuzzy incidence matrix between process modules; converting and solvingthe fuzzy incidence matrix by using a fuzzy clustering analysis method, and obtaining a transfer closure matrix; and forming a module clustering graph according to different partition threshold sequences, obtaining different module division schemes by selecting different lambda values. The complex process module division method is tight in logic, and is accurate and efficient in control, thus rapidly and efficiently producing personalized finished products adapting to market and technology changes, and improving the high-level requirements of the process.

The invention relates to a graph depth clustering method and device based on double-correlation reduction, computer equipment and a storage medium. The method comprises the following steps: acquiring a to-be-clustered original target relation graph; performing data enhancement on the original target relation graph based on the feature similarity and graph diffusion to obtain a classification graph adjacency matrix and a diffusion matrix, obtaining a first augmented graph according to the classification graph adjacency matrix and Gaussian noise, and obtaining a second augmented graph according to the diffusion matrix and the Gaussian noise; respectively inputting the feature maps of the first augmented map and the second augmented map into a pre-trained double-correlation reduction network to carry out correlation reduction of a sample level and correlation reduction of a feature level, and obtaining a double-correlation reduction feature map; and performing graph clustering on the double-correlation reduced feature graph to obtain a target relation clustering graph. By adopting the method, collapse can be avoided, and the clustering performance is improved.

The invention discloses a method for quickly quantitative grading of large-scale heterogeneous cluster nodes at a cloud data center, and the method logically comprises three main parts: cluster node performance parameter preprocessing, cluster node performance parameter matrix calibration, and cluster node performance parameter soft clustering. The method comprises the steps: firstly carrying out the collection quantification and standardization processing of performance parameters of different dimensions of cluster nodes at the cloud data center; secondly carrying out the calibration of the cluster node performance parameter values after standardization, introducing a similarity coefficient method based on the calibration values, and building a performance parameter fuzzy similarity matrix of all cluster nodes at the cloud data center; reconstructing the obtained fuzzy similarity matrix based on a transitive closure method, enabling the fuzzy similarity matrix to become a fuzzy equivalent matrix, carrying out the intercepting of the fuzzy equivalent matrix at a proper intercept level, and finally obtaining a large-scale cluster node performance parameter cluster graph. The method provides a node performance reference basis for the subsequent data layout and energy consumption management of the cloud data center.

The invention provides a user position prediction framework based on clustered graph federal learning. The user position prediction framework comprises the following steps that S1, using a sequence prediction model to carry out training locally by users; S2, uploading the model parameters and the implicit state of the original sequence data passing through an encoder to a server by users; S3, learning a similar graph structure by using an implicit state; S4, obtaining an embedded representation of the users through a graph convolutional neural network; S5, dividing the users into a plurality of clusters through a clustering method, and executing a federated average algorithm by the users in each cluster; and S6, downloading the embedded representation and the averaged model parameters to the corresponding users, splicing the implicit state and the embedded representation by each user, then outputting a prediction result, and updating the server model parameters. The method has the advantages that federal learning protects data privacy; the graph convolutional network solves the problem of insufficient training cost caused by label scarcity; and the graph clustering algorithm enables more similar users to execute a federated average algorithm so as to solve the problem of heterogeneity between the users.

The invention provides a detection method of abnormal data of an embedded intelligent terminal, and the embedded intelligent terminal. The detection method comprises: receiving data to be tested, wherein the data to be tested comprises the abnormal data; determining an object in the data to be tested, a k neighbor of each object and a k neighbor relationship between the objects; constructing a k neighbor directed graph according to the k neighbor relationship between the objects; constructing a k-clustering graph according to the k neighbor directed graph, and marking a clustering cluster in the k-clustering graph; and when a part of clustering clusters in the clustering cluster is meet preset conditions, outputting the data to be tested in the part of clustering clusters as the abnormal data. By adoption of the method, small-scale abnormal data mining can be performed in the environment of the embedded intelligent terminal by using a graph clustering method in the case that the computing resources of a terminal device are limited, the detection efficiency is high, the operation efficiency is high, and meanwhile the false alarm rate is reduced.

In order to overcome the defects of low clustering precision and weak robustness in clustering segmentation by adopting a least square regression method, the invention provides a manifold learning andrank constraint-based clustering method, which comprises the following steps of: obtaining original data, preprocessing the original data, and constructing a feature matrix X of the original data; based on a k-nearest neighbor method, calculating similarity among elements in the feature matrix X by adopting a similarity measurement function to obtain a weight matrix W corresponding to the featurematrix X; taking the weight matrix W as an initial matrix of a low-rank representation matrix Z, solving the low-rank representation matrix Z through a least square regression method, and applying manifold constraint and rank constraint to the low-rank representation matrix Z to obtain a final target function; and converting the final objective function from a constrained problem to an unconstrained problem by adopting a Lagrange multiplier method, alternately iteratively optimizing variables in the final objective function until convergence to obtain an optimal low-rank representation matrixZ ', and then obtaining a clustering result by adopting a spectral clustering graph cutting method for the optimal low-rank representation matrix Z'.

The invention discloses a large-scale data group searching method based on time sequence density clustering, which comprises the following steps: on the basis of collection of hundred million nodes online, carrying out preliminary pretreatment on the collected nodes, constructing an original cluster and a clustering graph for expressing a node relationship, and finding a group where a merging node is located according to a communication relationship of representative cluster nodes. Along with each round of iteration of the algorithm, node contribution degree scores at different moments are calculated, and range query is executed on the nodes according to the score values. Under the condition that the correctness of final group discovery is ensured, the search efficiency of a large-scale data network can be well improved by adopting a time sequence-based density clustering method. By the adoption of the scheme, the key problems of high-energy physical data intensive access and diversified data query requirements can be solved by reducing I/O and cross-domain communication traffic.

The invention provides a method for measuring and calculating the dispersity of a load type metal catalyst based on an atomic resolution electron microscope. The method specifically comprises the following steps: (1) acquiring a picture of the metal catalyst by adopting a transmission electron microscope; (2) identifying metal particles by carrying out hat transformation, substrate filtering by afrequency domain method, image sharpening, dynamic threshold binarization and target region marking sequentially; and (3) calculating the dispersity, by identifying metal particles in all pictures according to the step 2), obtaining a single-point graph and a cluster graph for each picture, calculating the dispersity according to picture information provided by the single-point graph and the cluster graph, and fitting dispersity into a function. The picture information comprises the number of metal particles, distribution positions of the metal particles, the number of clusters, the area of clusters and the area of part with a substrate of an image. The method of the invention can solve the problem that the prior art lacks a method to calculating and measuring the dispersity aimed at the catalyst in the atomic scale.

The invention relates to an urban traffic flow prediction method based on a Markov clustering graph attention network, and the method comprises the following steps: 1, obtaining a time sequence flow matrix according to historical transportation data; 2, natural structure information existing in the graph is extracted based on a Markov clustering algorithm idea, and a global correlation node matrix is obtained; 3, establishing a generative adversarial neural network model, and expanding an improved graph attention module in a model generator to global correlation node information obtained based on a Markov clustering algorithm instead of limiting neighbor nodes in first-order neighbor nodes in a graph attention network when acquiring spatial hidden features; learning the training model, and taking the learned model as a regional traffic flow prediction model; the improved graph attention module not only pays attention to local neighbor nodes, but also dynamically considers neighbor node information in an overall graph structure, and gives different weights to the neighbor nodes, so that the acquisition capability of spatial features is improved.

The invention discloses an optical network routing and wavelength allocation method and system based on a graph theory, and a storage medium. The method comprises the following steps: establishing a cluster graph model according to an optical network; carrying out sparse processing on the cluster graph model; carrying out initial division coloring on the cluster graph subjected to the sparse processing to obtain an initial solution; improving the initial solution to obtain an optimal solution; and outputting the optimal solution, generating a networking scheme according to the optimal solution, and completing the construction of the optical network. According to the scheme of the invention, the optical network routing and wavelength allocation problems are modeled by using a graph partitioning and coloring technology; a one-to-one correspondence relationship is formed, a routing and wavelength allocation scheme is designed according to a graph division coloring algorithm; compared withthe prior art, the scheme also has high efficiency for a larger network, not only can the solution quality be ensured, but also the efficiency can be ensured, so that the utilization of wavelength resources is effectively saved, a reasonable networking scheme is provided, and the scheme can be widely applied to the technical field of optical network transmission.

The invention discloses an image library construction method and device, an image retrieval method and device, computer equipment and a storage medium. The method comprises the steps that an image high-dimensional feature set is acquired; performing clustering processing on all high-dimensional features until a loss function converges to obtain at least two clusters; constructing an inverted file system based on all clusters; for each cluster, carrying out edge connection processing on all points in the cluster to obtain a complete graph corresponding to the cluster, and taking all the complete graphs as a current clustering graph; performing clustering processing on the current clustering graph to obtain at least two cluster centers; constructing an intermediate graph based on all cluster centers; taking the intermediate graph as a current clustering graph, returning to the step of carrying out clustering processing on the current clustering graph to obtain at least two cluster centers, and continuously executing until the number of layers corresponding to the intermediate graph reaches a preset number of layers, so as to obtain a multi-layer skip list; and constructing an image library based on the inverted file system and the multi-layer skip list. By adopting the method and the device, the image retrieval accuracy is improved.