59 results about "Graph Edge" patented technology

Provided is a process including: obtaining a first graph comprising nodes and edges, each of the first-graph edges linking two of the first-graph nodes and denoting semantic similarity of unstructured text in documents corresponding to the two linked first-graph nodes; for each of the first-graph nodes, selecting nodes for a second graph from attributes of the unstructured text documents to which the first-graph node corresponds, wherein the attributes are entities mentioned in the unstructured text documents, and wherein each of the second-graph nodes corresponds to a respective selected attribute; and for each pair of the second-graph nodes, determining a respective edge weight indicating similarity between a first entity corresponding to a first node of the respective pair and a second entity corresponding to a second node of the respective pair.

The invention discloses a method for eliminating a colorful edge of a display graph and a display device. The method for eliminating the colorful edge of the display graph comprises the steps of obtaining multiple edge pixels corresponding to a display graph edge in a display screen, respectively obtaining first pixels and second pixels which are in the same line with the edge pixels and are adjacent and third pixels and fourth pixels which are in the same column and are adjacent, adopting the edge pixels corresponding to the first pixels, the second pixels, the third pixels and the fourth pixels to perform compensation algorithm processing so as to obtain gray-scale values after the edge pixels are processed, adjusting the gray-scale values of the multiple edge pixels to be corresponding processed gray-scale values. By the adoption of the technical scheme, the advantage of eliminating the colorful edge of the display graph in the display screen is achieved.

The invention belongs to the technical field of network security and particularly relates to an intelligent defense decision-making method and device based on reinforcement learning and attack and defense games. The method comprises the steps that an attack and defense game model is constructed under finite constraint, an attack and defense graph used for extracting the network state and the attack and defense action in the game model is generated, the attack and defense graph is set to take a host as the center, attack and defense graph nodes extract the network state, and the attack and defense graph edge analyzes the attack and defense action; and when the network state transition probability of the defender is unknown, the defender obtains the defense revenue through online learning, so that the defender automatically makes the selection of the optimal defense strategy when facing different attackers. The game state space is effectively compressed, and the storage and operation expenditure is reduced. Adefender performs reinforcement learning according to environmental feedback in confrontation with an attacker, and can adaptively make an optimal choice when facing different attacks; therefore, the learning speed of the defender is improved, the defense income is improved, the dependence on historical data is reduced, and the instantaneity and the intelligence of the defender during decision making are effectively improved.

A method and system are provided. The method includes storing, in a memory, temporal data for a temporal graph. The memory includes a temporal graph storage structure having a set of buckets. The temporal data stored in the buckets includes respective data segments implemented using graph edges. Each of the graph edges has a start time and an end time associated therewith. The method further includes the methods, performed by a processor, of: forming an index that categorizes the graph edges based on the end times of the graph edges; positioning the graph edges within respective ones of the buckets for storage using the index such that the graph edges are positioned in the respective ones of the buckets in a chronological order that is based on the end times; and accessing the temporal graph storage structure using the index responsive to a temporal graph query.

A method for determining an optimal labeling of pixels in computer vision includes modeling an image by a graph having interior nodes and edges where each image point p is associated with a graph node, each pair of nearest neighbor points p, q is connected by a graph edge, each graph node p is associated with a singleton potential c(p), and each graph edge is associated with a pairwise potential function d(p,q). A label is randomly assigned to each point to initialize unary variables including an indicator function that indicates which label is assigned to which point and dual variables including height variables associated with each node p and label a, and balance variables associated with each edge (p,q) and label a. For each label, a new label c is selected, a capacitated graph is constructed and solved. The label selection divides the image into disjoint regions.

The invention relates to systems and methods for navigating, outputting and displaying non-linearly connected groups of transmedia content. Specifically, the invention involves retrieving, from a database, an ordered group of transmedia content data objects comprising a plurality of transmedia content data objects and linking data, whereby each element of the linking data defines a directional link from one of the transmedia content data objects to another of the transmedia content data objects; generating a two-dimensional graph structure representing the ordered group of transmedia content data objects whereby each graph node corresponds to a transmedia content data object and each graph edge corresponds to an element of the linking data; and outputting the graph structure on a display screen.

The invention discloses a recommendation method for graph construction and L1 regular matrix decomposition combination learning. The method comprises the following steps that: (1) according to user scoring data, establishing a scoring matrix; (2) on the basis of an L1 regular matrix decomposition model, taking graph edge weight as an unknown variable to be embedded into a target function so as to establish a new model; (3) on the basis of a score or a tag or a topic model, calculating an initial similarity between objects, and describing the similarity by a graph to obtain the initial graph edge weight; (4) randomly initializing a latent semantic feature vector to be solved in a decomposition model; (5) on the basis of the initial graph edge weight, using an OWL-QN (Orthant-Wise Limited-Memory Quasi-Newton) algorithm to adaptively update parameters in the decomposition model; and (6) obtaining an integral scoring matrix, and providing recommendation for a user according to the scoring matrix. By use of the method, through combination learning, the effective graph edge weight and the latent semantic feature vector can be obtained through the combination learning so as to obtain good prediction presentation.

A method for the creation of rectilinear Steiner minimum trees includes determining a set of candidate connections from a terminal node to a different terminal node or to a graph edge. The length of each candidate connection may be used to determine the set of candidate connections that span the graph with a minimum total length.

The invention provides a variable weight cost aggregation stereo matching algorithm based on a horizontal tree structure. The method comprises the stages of calculating matching cost, calculating a guide image, constructing a horizontal weight tree and a vertical weight tree, cost aggregation, updating the weight trees, cost aggregation and disparity refinement. The initial disparity value is acquired in the cost aggregation stage and then the weight of four domain graph sides is correspondingly adjusted, the smooth constraint and weight adjustment are combined and a new horizontal weight tree and a vertical weight tree are constructed by updating the weight of the four domain graph sides, and aggregation is performed along the horizontal tree and the vertical tree and the cost aggregation result indicates that the matching accuracy for the background region and the image boundary region can be enhanced to a certain extent.

A graph database is described. The graph database includes one or more symbolic data stores and one or more key-value data stores. Each symbolic data store is configured to symbolically store sets of multiple hyper-graph nodes. Each key-value data store is configured to store attribute information for hyper-graph nodes and hyper-graph edges.

The invention belongs to the technical field of knowledge graph application, and particularly provides a network fraud identification method and system based on a knowledge graph, and the method comprises the steps: constructing a mode layer of a network fraud knowledge graph from top to bottom; loading data to the data layer of the network fraud knowledge graph; based on the constructed network fraud knowledge graph, fusing a third-party knowledge base in a data layer, and expanding the data volume of the data layer of the network fraud knowledge graph; discretizing and normalizing attribute frequency values of key behavior relationships in the network fraud knowledge graph, and converting the attribute frequency values into weight values of graph edges; and based on a network fraud case graph, by utilizing a spark graphics graph and an LPA algorithm, identifying a criminal gang, and then by utilizing a PageRank algorithm, identifying core members in the gang. The network fraud case and the criminal gang can be effectively organized through entities, relationships and attributes by using the strong relevance of the knowledge graph. The case detection efficiency is improved, and the occurrence rate of network fraud cases is reduced.

The invention relates to a three-dimensional sketch structure design method and device, computer equipment and a storage medium. The method comprises the steps that a three-dimensional sketch structure used for storing streaming data information is constructed, each sketch in a three-dimensional sketch array is divided into three areas, the streaming data information of a graph structure is obtained, and the streaming data information is updated into the three-dimensional sketch structure according to hash functions corresponding to different depths. According to the method, representative keyreservation and majority voting are combined, a constructed three-dimensional sketch structure maintains structure information and weight information of graph data at the same time, once stream datais updated only once, a constant probability error range is guaranteed, stored keys are selected by using a majority voting algorithm so as to record most representative graph edges, and the most representative graph edges are stored. An error range may be reduced and reversibility may be achieved to improve query efficiency.

The invention provides an image stitching processing method, a device, and an electronic system. The method comprises the steps of obtaining a first overlapping image and a second overlapping image corresponding to overlapping areas of a first image and a second image to be stitched respectively; determining a fusion energy diagram corresponding to the first overlapping image and the second overlapping image; determining a minimum energy path corresponding to the first graph edge to the second graph edge of the fused energy graph; determining a first mask corresponding to the first overlapping image and a second mask corresponding to the second overlapping image based on the minimum energy path; and stitching the first image and the second image based on the first mask and the second mask. According to the invention, the area through which the abutted seam passes is optimized to a certain extent, and the image splicing effect is improved.

The invention provides a preprocessing method for an optical proximity correction technology, and belongs to the technical field of integrated circuit design. The method comprises the following steps that: enabling a first graphic edge to extend to the outer side of a graph for a first preset distance so as to obtain a first rhombic graph; enabling a second graphic edge to extend to the outer side of the graph for a second preset distance to obtain a preprocessed coverage graph, and enabling the preprocessed coverage graph to extend a direction parallel with the second graph edge for a third preset distance to obtain a coverage graph which covers parts of first graph; carrying out logic ''NOT'' processing on the first graph and the coverage graph to obtain a second graph; carrying out logic "Or" processing on the graph and the second graph to obtain a first target graph with a chamfering notch; and carrying out triangular filling on the chamfering notch to obtain a second target graph. The preprocessing method has the beneficial effects that the accuracy and the consistency of an OPC (Optical Proximity Correction) processing result are improved.

A graph display system that represents graph nodes as grid cells in a two-dimensional grid. Graph edges are represented by placing the edge's nodes in adjacent grid cells, thus eliminating the entanglement of edges that are typically drawn as lines between scattered nodes. Relatedness of two nodes sharing an edge is indicated by the edge's weight. The display system prioritizes placement of higher weight edges such that more related information is grouped together in nearby grid cells. Any graph with n nodes and e edges can be displayed and browsed.

The invention discloses a distributed graph calculation method, a terminal, a distributed graph calculation system and a storage medium. The embodiment of the invention is based on graph data preprocessing, a graph division algorithm and an incremental graph edge sorting algorithm. According to the method, the original graph data is converted into the computer readable intermediate graph data, sothat subsequent graph division can be performed quickly, high-quality graph division is provided, the communication overhead of the generated high-quality partition is greatly reduced, and the calculation and analysis speed of the distributed graph is increased.

The invention discloses a room segmentation method based on internal circles and an adjacency graph, and the method comprises the steps: carrying out distance transformation on a two-dimensional occupancy probability grid map obtained through laser scanning, and calculating the distance value of a nearest occupancy point of each pixel; filling a free space with internal circles, and constructing internal circle approximation of the indoor free space; constructing an indoor topology network based on the adjacency relation of the internal circles, and constructing the incidence relation between the internal circles by a weighted undirected adjacency graph; adding and deleting undirected graph edges based on rules, and achieving connected sub-graph segmentation of the graph; and combining the internal circle connected sub-graphs to obtain different adjacent circle clusters, endowing the adjacent circle clusters with different room semantic information, and finally realizing semantic segmentation of an indoor free space (namely, a room). The method can be applied to room segmentation and task planning of the indoor space in the laser scanning process of a mobile robot.

The invention discloses a banknote denomination identification method and device. The banknote denomination identification method mainly comprises a training part and an identification part. The training part is characterized in that feature extraction can be carried out automatically after enough samples are collected, human intervention is not needed, and the recognition efficiency is greatly improved. The identification part is characterized by comprising the following steps: acquiring graph edge information; finding graph coordinates through edge information fitting, and performing plane transformation according to the coordinate information to obtain transformed graph coordinates; obtaining and calculating one-dimensional feature information of the image according to a preset method;and matching with one-dimensional feature information of a standard image, and judging whether the matching feature is abnormal or not. According to the invention, the feature two-dimensional featuresare converted into a plurality of one-dimensional features, the denomination one-dimensional features are calculated, and the calculated one-dimensional features are matched with the standard denomination one-dimensional features. The invention does not depend on the image type, is not affected by the CIS type and the difference between the new bank note and the old bank note, and has high adaptability to old bank notes, residual bank notes and stained bank notes.

The invention provides a two-dimensional Euler diagram generation method and device for a one-stroke game, and the method comprises the steps: generating points meeting all parameter values accordingto all input parameter values of a two-dimensional Euler diagram; carrying out random distribution on each point to generate an initial two-dimensional Euler diagram edge set; obtaining the degree ofeach point, and adding a new edge to the edge set which does not meet the degree of the one-stroke game point; modifying a connection graph in the edge set after the new edge is added, and determininga modified two-dimensional Euler diagram; symmetrically overturning the modified two-dimensional Euler diagram to obtain a final two-dimensional Euler diagram; and generating a stroke sequence on thefinal two-dimensional Euler diagram by using a Flory algorithm. Through parameter input of the two-dimensional Euler diagram, generation of two-dimensional coordinates of all points and generation ofan edge set meeting the degree limitation of the two-dimensional Euler diagram points, the final two-dimensional Euler diagram is obtained, the attractiveness is improved, public aesthetic appreciation is met, and users can accept the two-dimensional Euler diagram more easily; meanwhile, the content generation efficiency of one-stroke games is improved, and the yield is increased.

The invention discloses an OPC method for a specific graph side wave effect, which comprises the following steps of S1, setting an exposure threshold of an OPC model, inputting a target graph, simulating light intensity distribution of the exposed target graph by utilizing the OPC model, selecting a rectangular graph of which the light intensity meets a set light intensity condition as a selected graph, and outputting the selected graph, S2, taking the two long edges of which the distance between the two long edges in the selected graph meets a set distance condition as graph edges, S3, taking the graph edge as a target graph edge, generating a scattering strip in the selected graph, and enabling the scattering strip to be parallel to the graph edge, and S4, with the target graph as a target and the scattering strips as auxiliary graphs, conducting model-based OPC correction to obtain a final OPC graph. The invention further discloses an OPC processing method for the through hole layer. The light intensity distribution is changed by adding the auxiliary patterns (scattering strips), so that the final exposure result can reach the target size, and the generation of extra exposure patterns can be avoided.

The invention discloses an online multi-target tracking algorithm based on graph matching. A data association problem of detection response between two continuous frames is converted into a graph matching problem. Firstly, two deep convolutional neural networks are designed to respectively solve the intimacy between vertexes of two images and the intimacy between edges of the two images; then theintimacy between the vertexes and the intimacy between the edges are directly used to fill the intimacy matrix between the two images, and finally the intimacy matrix is processed to obtain a final matching matrix (i.e., an incidence matrix between detections). Therefore, the correlation of real data in the multi-target tracking process can be effectively reflected, and the accuracy of the tracking result is high.

The invention relates to an image data processing method, a pixel image obtained by the image data processing method and an application system. According to the method, graph pixels are decomposed into five basic elements, namely black points, white points, geometric points, semi-angular points and full-angular points, and then data processing and generation are carried out. The method comprises the following steps: 1, classifying graphs; 2, carrying out proper region division on the graphs; 3, respectively processing and calculating the size and the pixel density of graph zone bit to form zone bit pixel sorting and data logic relationship processing; 4, performing comprehensive calculation to obtain a basic pixel graph and a pixel application data value; 5, processing external data by using the basic pixel elements for editing, importing each area bit graph, and combining and encrypting to form a main storage area; and meanwhile, color, sound or 3D graphic processing is added according to design requirements. 6, scanning and analyzing rows and columns after the main storage area is filled, and completing automatic positioning and error correction; 7, forming a graph edge boundarythrough operation to distinguish a main storage identification area; 8, clearing pixel error codes and internal and external messy codes of the graph; and 9, outputting a complete pixel graph.

The invention belongs to document layout comprehension technology in intelligent text and graphic information processing, in particular to a content-based method for recovering text reading order of newspaper layout. Aiming at the disadvantages of losing the reading order and not having chapter independence in dealing with complex newspaper layouts in the prior art, this invention uses graph theory to mathematically model this problem for the first time, expressing the adjacency relationship of text blocks as a directed graph, and Split the directed graph into a weighted bipartite graph, use natural language processing technology to calculate the edge weight of the bipartite graph, and obtain multiple continuous sequences through optimal matching, and then divide each sequence into multiple subsequences according to the style information of the text block. The connection of the content corresponding to the sequence is the text flow of the independent chapters with the reading order. Utilizing semantic, spatial relation and style information, the correct rate of reading order recovery is greatly improved and independent of chapters. This method can be applied to layout understanding and structural reconstruction of styled documents.

The invention discloses an online multi-target tracking algorithm based on graph matching, which transforms the data association problem of detection responses between two consecutive frames into a graph matching problem. Firstly, two deep convolutional neural networks are designed to solve the intimacy between the vertices of the two graphs and the intimacy between the edges of the two graphs respectively, and then directly use the intimacy between the vertices and the intimacy between the edges to fill the two The intimacy matrix between two graphs, and finally the intimacy matrix is ​​processed to obtain the final matching matrix (that is, the correlation matrix between detections). Therefore, the present invention can effectively reflect the correlation of real data in the process of multi-target tracking, and the accuracy of tracking results is high.

The invention provides a topological structure generation method and system, a topological structure updating method and system and a storage medium. The generation method is applied to a network system constructed on the basis of a graph structure, and comprises the following steps: according to a preset layering condition of an nth layer, performing topology on the basis of a (n-1) th layer layered forest by utilizing graph edges in the graph structure to obtain the nth layer layered forest, wherein the layered forest of the zeroth layer is empty, and n is greater than or equal to 1; progressively increasing the n value from 1 and repeating the above steps until all points in the graph structure are traversed in a layered forest formed in the h layer to obtain a spanning tree of the graph structure, wherein n < = h, the layering condition of the nth layer is set based on the tree degree of a point in the graph structure in the (n-1) th layer and the tree degree difference of the point between the (n-1) th layer and the nth layer. According to the topological structure generation method, the minimum spanning tree which is beneficial to updating and has low resource consumption can be constructed on the basis of a graph structure.

The invention provides a method for constructing and displaying a multimodal emotional knowledge graph, and relates to the technical field of multimodal knowledge management. This method first collects and downloads emotional vocabulary ontology or dictionaries, facial expression and animal expression images, and emoticons commonly used on social media platforms; and preprocesses the acquired text and image data resources respectively, so that these data resources meet a unified Abstract ontology model requirements; then based on the preprocessing results of text emotional words, emoticon images, and social media emoticons, construct a multi-modal emotional knowledge map that can express different modal entities and the relationship between entities; finally, by using front-end text tags The language and the programming of the element rendering language of the front-end page are used to realize the display and rendering of the nodes, edges of the graph, and attribute information of the graph entity in the multi-modal emotional knowledge graph; on this basis, the user interaction function of the graph is realized; at the same time, through the graph The visualization library increases the dynamic effect of the knowledge map.

The invention relates to a graph common-edge detection method, device and equipment and a computer readable medium. The method comprises the steps: acquiring a first graph and a second graph, whereinthe first graph is a to-be-edge-shared graph, and the second graph is a reference graph; performing target Boolean operation on the first graph and the second graph; and determining a first edge between the first graph and the second graph according to an operation result of the target Boolean operation, wherein the first edge is a common edge of the first graph and the second graph. According tothe invention, Boolean operation is carried out on the graphs to judge whether the two graphs have a relationship or not and change the graphs needing edge sharing so as to achieve accurate labeling of the intersection part between the objects in the images, so that the accuracy of the graph edge sharing result is improved, and the graph edge sharing detection process is simplified.

Infrastructure attacks based on graph edge context are identified by receiving an execution graph constructed by a central service based on behaviors monitored by a plurality of agents deployed on respective systems including a first system. The execution graph comprises a plurality of execution trails. One or more tags are applied to each edge of an execution trail of the execution graph based on at least one of temporal context or spatial context associated with the edge. One or more behaviors associated with the edge of the execution trail happen across an enterprise infrastructure involving the first system. The execution trail enriched with the one or more tags is analyzed. An action that is performed to mitigate security risks in the execution graph is determined based on the analysis.