66 results about "Graph similarity" patented technology

The invention belongs to the mobile Internet technical field and particularly relates to a class dependency graph based Android application similarity detection method. The class dependency graph based Android application similarity detection method specifically comprises step 1, decompiling an Android application and obtaining Dalvik byte codes of the Android application; step 2, obtaining package names and class names in the Android application according to a decompiled file directory and Dalvik byte code file names of the Android application; step 3, establishing a dependency relation graph between classes according to the package names, the class names and Dalvik byte code files; step 4, repeating the step 1 to the step 3 until class dependency graphs of Android application which need to be compared are obtained; step 5, performing comparison on similarities of the Android applications by a graph similarity comparison method according to the class dependency graphs of the Android applications. According to the class dependency graph based Android application similarity detection method, the structured information in the Android applications can be effectively extracted and the influences on the detection rate of the similar applications from an Android application confusion and deformation technology can be resisted.

The invention discloses an anomaly detection method based on data incremental graphs. The anomaly detection method includes the following steps that detection data in a current monitoring zone of a wireless sensor network are collected and preprocessed, and an event zone is determined; data sets relevant to a current event are acquired, a graph model is utilized to abstractly generalize event data, and the event data are converted into the event data incremental graphs; a graph similarity algorithm based on structure correlation is utilized to search an event mode graph database for event mode graphs similar to the event graphs and judge the type of the current event, wherein the event mode graph database is a set of the event mode graphs; the event mode graphs are the event data incremental graphs and abstract description for types of events; by the adoption of the graph similarity query algorithm based on the structure correlation, the graph similarity query problem is converted into the sequence similarity query problem, and therefore query complexity is effectively reduced. By the adoption of the anomaly detection method based on the data incremental graphs, the event graphs can be acquired based on domain expert knowledge or data analysis and used for detecting complex events, the detection efficiency of the events is improved, and the false alarm rate is reduced.

The invention discloses an error correction and compensation method and device based on a calibration board in a direct-writing exposure machine. The method includes the steps: firstly, acquiring vertex coordinate values of at least one arbitrary area on the calibration board; secondly, determining size expanding and shrinking variation quantity of the area corresponding to the vertex coordinate values according to a graph similarity transformation model and the vertex coordinate values; finally, performing error correction and compensation for a direct-writing exposure machine system according to the size expanding and shrinking variation quantity. According to the method, deformation of graphs of different areas is measured and corrected by the aid of the calibration board, correction and compensation of an exposure device system are achieved, so that the quality exposing imaging results of different areas are improved.

Techniques are disclosed relating to malware clustering based on function call graph similarity. In some embodiments, a computer system may access information corresponding to a plurality of malware samples and, based on the information, generate a function call graph for each of the malware samples. In some embodiments, generating the function call graph for a given malware sample includes identifying a plurality of function calls included in the information, assigning a label to each of the function calls, identifying relationships between the function calls, and generating the function call graph based on the relationships and the labels. Based on the function call graphs, the computer system may assign each of the plurality of malware samples into one of a plurality of clusters of related malware samples.

The invention relates to the technical field of image detection, and discloses a graph similarity judgment method, which comprises the following steps of performing semantic segmentation on a target picture needing similarity comparison to obtain a target content label corresponding to the target picture after semantic segmentation; searching a known content tag database corresponding to known image characteristics according to the obtained target content tag, and identifying whether a known content tag which is the same as the target content tag exists in the database or not; and if the knowncontent tag which is the same as the target content tag exists, obtaining a known picture corresponding to the known content tag, carrying out image texture similarity comparison on the target picture and the known picture, and determining whether the target picture is similar to the known picture or not according to a comparison result. The invention further provides a graph similarity judgmentdevice and a computer readable storage medium. According to the method, the recognition efficiency and accuracy of the graph similarity are improved, and the retrieval efficiency and reliability of graphs are improved.

The invention discloses a computer method for comparing similarity of different plant forms. The plant topological structure features, peripheral contour features and organ geometry structure information are utilized, and the method comprises the steps of firstly adopting the simplified topological structure of a tree graph describing plant, then computing the similarity of the topological structure based on a tree graph editing distance method, then computing the similarity of three-dimensional convex hulls comprising plants among projections of a plurality of planes based on the two-dimensional graph similarity algorithm, on the basis of the similarity, integrating two-dimensional similarity values in all the projections so as to compute the similarity of peripheral contours of the plants, then, comprehensively considering the geometric attribute of subtrees of all levels, computing the similarity of geometric details of the plants, finally, fusing multi-feature information of the plants, and calculating the similarity among different plant structures. According to the method, form similarity of the different plants can be quantized, and the aim of distinguishing varieties or subjects of the different plants through the similarity can be achieved.

In order to meet the engineering demand that targets in a column are accurately tracked through multiple sensors under the complicated background of cloud and rain clutter, banding interference and the like and overcome the defect that a traditional multi-sensor multi-target tracking algorithm and an existing column target tracking algorithm both have difficulty in achieving an ideal tracking effect, the invention provides a centralized multi-sensor column target particle filtering algorithm based on shape and direction descriptors according to the characteristic that true echo space structures of the targets in the same non-motorized column at adjacent moments are fixed relatively. According to the centralized multi-sensor column target particle filtering algorithm based on the shape and direction descriptors, state update of multi-dimensional trace points of the targets in the column is achieved based on particle filtering according to redundancy trace points in graph similarity removal state prediction, and the targets in the centralized multi-sensor column are tracked accurately.

The invention discloses an anomaly detection method based on data snapshot graphs. The method includes the first step of carrying out acquisition and pretreatment on detection data in a current monitored area of a wireless sensor network to determine an event area, the second step of obtaining a dataset related to a current event, using a graph model to abstractly summarize event data and converting the event data into the event data snapshot graphs, and the third step of carrying out query in an event mode pattern database through a graph similarity algorithm based on structural correlativity, searching for event mode patterns similar to the event graphs and judging the type of the current event, wherein the event mode pattern database is a collection of the event mode patterns, and the event mode patterns are the event data snapshot graphs which represent for abstract description of the type of the event. According to the anomaly detection method based on the data snapshot graphs, the event graphs can be obtained on the basis of domain expert knowledge or on the basis of data analysis. The method has the advantages of being used for detection of the complex event, improving event detection efficiency and reducing the false alarm rate.

The invention provides a similar medical record searching method, device and equipment and a readable storage medium. The method comprises the steps: obtaining query medical record data and multiple pieces of historical medical record data; obtaining query graph structure data corresponding to the query medical record data and historical graph structure data corresponding to the historical medicalrecord data, wherein the query graph structure data and the historical graph structure data both comprise first-class sub-graphs and second-class sub-graphs, and intermediate nodes and leaf nodes ofthe second-class sub-graphs are obtained by conducting feature recognition on the first-class sub-graphs; according to the root node similarity, the first type of sub-graph similarity and the second type of sub-graph similarity, obtaining the similarity degree of each historical graph structure data and the query graph structure data; according to a preset selection rule and the similarity degree,determining a similar medical record search result for querying the medical record data, so that inherent and recognizable sub-graphs in the medical record data are extracted, the relevance of corresponding sub-graph contents is measured in comparison, and the similar medical record searching accuracy is improved.

Embodiments of the invention provide a data processing method, device and equipment and a medium, and the method relates to an artificial intelligence technology, can be applied to the field of natural language processing, and comprises the steps of obtaining a target text and a standard text, and generating a target entity sub-graph corresponding to the target text and a standard entity sub-graphcorresponding to the standard text according to a knowledge graph, wherein a first entity in the target entity subgraph and a second entity in the standard entity subgraph belong to entities in the knowledge graph; according to the first entity and the second entity, generating a target graph structure feature corresponding to the target entity sub-graph and a standard graph structure feature corresponding to the standard entity sub-graph; and determining graph similarity between the target entity sub-graph and the standard entity sub-graph according to the target graph structure features andthe standard graph structure features, wherein the graph similarity is used for indicating the association degree between the target text and the standard text. By adopting the embodiment of the invention, the matching accuracy between the target text and the standard text can be improved.

The invention discloses an APT event homology judgment method based on a behavior pattern. The APT event homology judgment method comprises the following steps: constructing an APT event association graph based on the behavior pattern; performing node attribute expansion on the APT event association graph; performing attribute labeling on nodes in the APT event association graph; performing attribute labeling on nodes in the APT event association graph to obtain behavior tags or clue tags of the nodes of the APT event association graph; carrying out similarity discrimination on the APT event association graph; carrying out similarity judgment on the APT event association graph to complete homologous judgment of the APT event; and comparing topological structure information and node attribute information of the structural data of the two APT event association graphs by adopting a sub-graph similarity measurement function to determine the homology or similarity of the two APT event association graphs. According to the method, the problems of one-sided malicious sample analysis result and low manual homology judgment efficiency in existing APT event homology analysis are solved.

The similarity between graphs having an extremely large number of nodes, such as an SNS, a link of WWW, etc., can be obtained within a reasonable time. A unique value is provided to a label of a node in a graph. Preferably, the value is a fixed-length bit string. In this case, the length of the bit string is selected to be sufficiently larger than the number of digits by which types of labels can be expressed. With respect to one graph, nodes of the graph are sequentially visited by an existing graph search method, such as a depth-first search, a breadth-first search, and the like. At this time, in the system, when one specific node is visited, a calculation is performed for bit string label values of all nodes adjacent to the specific node and the bit string label value of the specific node, to obtain a bit string value. The hash calculation is performed for the calculated bit string value and the original bit string label value of the node to obtain another bit string label value, and this value becomes the label value of the node. After finishing the visit to all nodes in one graph, the label values of all nodes are rewritten. When the same treatment is performed for another graph which becomes a target of the graph similarity comparison, label values of all nodes in this graph are rewritten. Therefore, with respect to one graph, a ratio of label values which are identical to the label values in another graph, per all nodes is calculated to obtain the similarity.

The invention provides a closed graph similarity search method based on time sequence complexity difference, which comprises the following steps: S1, observing a closed graph to obtain a time sequenceof the closed graph; S2, setting the complexity difference CO of the two time sequences; S3, setting a complexity time measurement standard TSD; S4, finishing neighbor search on the complexity time measurement standard in the S3 by adopting an exhaustion method; S5, correcting the distance value by adopting a triangular inequality; S6, obtaining a closed position of the two-dimensional relative graph in the serialized index data structure; the problem that an existing graph similarity search method cannot achieve similarity search of two-dimensional graphs is solved. Meanwhile, the time dimension of a time sequence is greatly solved, the effectiveness of graph rotation in the measurement process is ensured, measurement standards are given for time sequences with different complexity degrees, and finally similarity search of two-dimensional closed graphs is achieved.

The invention discloses a method for enhancing trademark graph similarity judgment accuracy. Secondary retrieval and even multiple times of retrieval can be performed on the initial sorting result generated by the trademark arbitrary method and then final sorting is performed according to the set generated by final retrieval. According to the method, the sorting results of multiple times of retrieval of the correlation images are fused, the images having high appearance frequency have high weight, and the top sorted images have high weight so that the correlation between the images can be fully mined and the accuracy of trademark graph similarity judgment can be greatly enhanced. Meanwhile, the extraction and comparison mode enabling the initial sorting result to be more stable and accurate is also given for further enhancing the accuracy so as to provide a great foundation for subsequent secondary retrieval and even multiple times of retrieval.

The invention discloses a legal case similarity calculation method and system based on knowledge graph matching, wherein the method comprises the steps: converting entities and relation information in knowledge graphs of two cases into initial vectors through a text encoder, obtaining the vectorized representation of the entities in the knowledge graphs through a first graph neural network, and obtaining a matching score matrix between the two images according to a graph matching algorithm; and pooling the matching score matrix to obtain a first score vector, and inputting the first score vector into a multi-layer perceptron to obtain a similarity score of the two cases. The system comprises a text coding module, a knowledge graph embedding module, a matching score matrix calculation module, a matrix pooling module and a graph similarity calculation module. The method can be applied to different types of more complex cases; and local matching and global matching can be further planned as a whole, and a one-to-one correspondence relationship among graph nodes is approximately obtained, so that a law case similarity result with higher precision is obtained.

The invention relates to the technical field of computers, and discloses a fault self-recovery method and device, equipment and a storage medium. The method comprises the following steps: generating a time sequence system diagram according to system monitoring information collected regularly; if the fault is detected, extracting an abnormal sub-graph from the time sequence system graph at the current moment; and searching a fault recovery scheme matched with the abnormal sub-graph in a knowledge base by adopting a graph similarity algorithm, and automatically executing the fault recovery scheme. According to the technical scheme of the embodiment of the invention, the beneficial effects of realizing fault self-healing, improving the operation and maintenance efficiency and saving time and labor cost are achieved.

The invention discloses a line draft type graph similarity judgment method which comprises the following steps: a preprocessing step: obtaining a vector line draft image to be judged, carrying out graying processing and dual polarization processing on the vector line draft image to be judged, enabling the vector line draft image to be judged to present a black-white-gray state, and adaptively scaling the vector line draft image to be judged to a standard size; a comparison step: dividing the vector line draft image to be judged into n*n square areas to obtain a square matrix corresponding to the vector line draft image to be judged; and projecting the square matrix in the row direction and the column direction to obtain pixel point distribution conditions which are linearly distributed inthe row direction and the column direction, comparing the pixel point distribution conditions with vector line draft images of each standard size stored in an image library, and judging the similarityof the vector line draft images. The invention further discloses electronic equipment and a storage medium. The line draft graph similarity judgment method and device can obtain an ideal and accurateline draft graph similarity judgment result.

The invention discloses an image similarity judgment method, which comprises the following steps of: graying processing: obtaining a vector image to be judged, graying the vector image to be judged, enabling the vector image to be judged to present a black-white-gray state, and adaptively scaling the vector image to be judged to a standard size; A pixel point distribution comparison step, dividingthe vector image to be judged into n * m square areas to obtain a square matrix corresponding to the vector image to be judged, performing pixel point distribution comparison with the square matrix corresponding to each standard size image stored in the image library, and judging the similarity of the vector image according to the pixel point distribution condition of the square matrix of the corresponding specification. The invention further discloses electronic equipment and a storage medium, and the image similarity judgment method, the electronic equipment and the storage medium can obtain an ideal vector graph similarity judgment result, so that vector graph similarity judgment is more accurate.

The invention provides a visual pedestrian re-recognition method based on sparse graph similarity migration. The method comprises the steps: extracting features of each pedestrian image and a query image in a pedestrian image database through employing the same trained deep convolutional network, and representing the features through feature vectors; calculating the similarity of any two pedestrian images through the feature vectors, and constructing a database image dense association graph; performing sparse constraint on the database image dense association graph to obtain a database image sparse graph; using an energy minimum random walk model to take a value for calculating the similarity between the query image and the pedestrian image as an energy random migration in a database image sparse graph, taking a stabilized energy value as a consistency score of the query image and the pedestrian image, and sorting the database image sparse graph based on the consistency score to obtain a database image sparse graph; and returning the pedestrian image with the highest score. By the adoption of the scheme, the retrieval precision and speed of visual pedestrian re-recognition are improved.

The invention relates to the technical field of artificial intelligence and discloses a picture similarity calculation method and device, computer equipment and a storage medium. The picture similarity calculation method comprises steps of determining whether targets of the same type exist in a first picture and a second picture or not according to a first target detection result of the first picture and a second target detection result of the second picture; obtaining a first target sub-graph and a second target sub-graph from the first picture and the second picture when it is determined that the same type of target exists; calculating the sub-image similarity of the first target sub-image and the second target sub-image, determining the image similarity between the first image and the second image according to the sub-image similarity, and outputting the image similarity and the sub-image similarity; and when it is determined that the same type of target does not exist, performing edge detection on the first picture and the second picture, calculating picture similarity between the first picture and the second picture, and outputting the picture similarity. According to the method, accuracy of the similarity of the pictures can be improved, and similarity of the pictures can be explained.

The invention relates to a method and a device for corresponding a student with a driving training vehicle. The method comprises the steps of when the student conducts driving training each time, acquiring the driving training vehicle which is employed for practicing by the student each time according to a name in vehicle electronic identification carried by the student; preferentially arranging the driving training vehicle to the student for use in the case that the driving training vehicle is not employed; and if the driving training vehicle is employed, finding out a driving vehicle which is not used and has the highest similarity with the driving training vehicle by use of a SIFT (Scale-Invariant Feature Transform) graph similarity algorithm, and assigning the driving vehicle to the student for driving practice. Thus, when practicing driving each time, the student can be guaranteed to drive the same or similar driving training vehicle, and the time spent by the student for knowing the vehicle is shortened. The method employed by the invention can be implemented by a computer program stored in a computer readable storage medium through establishment of functional modules and combination of the functional modules into a functional module framework.

The invention discloses a method and a device for merging interface jump path diagrams. According to the method, similarity analysis is carried out on interface scenes on each interface jump path generated by testing the UI application interface by an automatic testing tool, and the similar interface scene nodes are merged, so that each interface jump path is merged, the merged interface jump path diagram is generated, and the complexity of subsequent analysis is reduced. According to the method, when the interface scene similarity is analyzed, a method of combining interface screenshot and control layout similarity comparison is adopted. Wherein when the similarity of the interface screenshots is calculated, feature vectors are extracted through a convolutional neural network, and then the distance between the feature vectors is used as the similarity of the interface screenshots; and when the control layout similarity is calculated, the control layout is converted into simplified layout texts, and then the text similarity between the layout texts is calculated.

The invention provides a calculation graph evolution AI model automatic generation method based on structural similarity. In the AI model automatic generation method based on calculation graph evolution, a graph similarity technology, namely a method of combining multiple graph similarity calculations, is adopted for calculating the similarity degree between a generated new network model and a known network model in the automatic generation process of an AI model, and a large number of repeated models, similar models and performance similar models are inhibited from appearing within a certainsimilarity threshold range. The diversity of model samples in the model search process can be effectively ensured, and the model network search success rate is improved. The performance degradation ofthe search network is obviously reduced. Meanwhile, the similarity threshold is dynamically adjusted according to the search efficiency, the function of jumping out of local optimum can be achieved according to model samples, and model search is accelerated. Therefore, the automatic generation efficiency of the AI model is improved.

The invention provides an inter-graph similarity calculation method and system, terminal equipment and a storage medium. The method comprises the following steps: respectively calculating a parameterized vector function of each polygon in the first graph and the second graph, pairing the corresponding polygons according to the parameterized vector function of each polygon in the first graph and the second graph, classifying the polygons which are successfully paired into a pairing set, and classifying the polygons which are not successfully paired into a non-pairing set; and performing distance calculation on all polygons in the pairing set and the non-pairing set to obtain a plurality of distance values, performing weighted calculation on the distance values to obtain a graph distance, and calculating graph similarity according to the graph distance. Through the design of calculating the parameterized vector function of each polygon, the graphic information of each polygon is mapped into the parameterized vector function, the calculation of the inter-graphic distance between two graphs under different shapes is facilitated, and the accuracy of inter-graph similarity calculation isimproved.

The invention belongs to the technical field of software development, and discloses a case graph similarity judgment method based on maximum common sub-graph calculation, which comprises the following steps of: 1, preprocessing to-be-compared UML (Unified Modeling Language) case graphs, and representing the UML case graphs as directed graphs; 2, calculating and obtaining a maximum common sub-graph between the to-be-compared directed graphs; and 3, calculating the similarity by using a similarity judgment algorithm. The maximum common sub-graph algorithm used in the method is simple in process, the graph structure is directly analyzed, the efficiency is high, it can be guaranteed that the use process is efficient and convenient, and the method has high applicability.

The invention discloses a hub model retrieval method. The hub model retrieval method comprises the following steps: receiving a to-be-retrieved first hub picture; acquiring first feature information according to the first hub picture; obtaining a first view of each hub model according to all hub models in a model database; acquiring second feature information of each hub model according to the first view; and sorting all hub models according to the first feature information and the second feature information. The invention further discloses a picture retrieval storage medium and equipment. According to the first feature information of the first hub picture, the second feature information of the first view is compared, the pictures in the hub model database are listed and retrieved, the process is simple, the accuracy of hub graph similarity judgment is improved, the calculated amount of the picture database is effectively reduced, and the purposes of saving cost and improving retrieval efficiency are achieved.

The embodiment of the invention provides a graph similarity analysis method, device and equipment and a storage medium. The method comprises the following steps: superposing the central point of a to-be-analyzed graph with the central point of a reference graph; determining a point in the overlapped area of the reference graph and the to-be-analyzed graph as a starting point; emitting more than a preset number of different rays from the starting point; determining the distance between a first intersection point of each ray and the contour line of the reference graph and a second intersection point of each ray and the contour line of the graph to be analyzed to obtain a distance set; determining the similarity between the to-be-analyzed graph and the reference graph according to the distance set; therefore, the time and the cost of similarity analysis are reduced through the similarity of the graph according to the distance between the two cross points.

The invention discloses a graph similarity calculation method and device based on a graph convolution network, can overcome the defect that an existing GNN-based graph similarity calculation model cannot well learn the hierarchical structure of a graph, and improves the effectiveness of graph similarity calculation by combining the flatness and hierarchical representation of the graph together within reasonable time. Specifically, the embodiment of the invention provides a function which completely supports back propagation and is based on an end-to-end ground neural network, each part of thefunction is carefully designed, so that the function can learn flat and hierarchical information of graphs, and finally, a pair of graphs is mapped into similarity scores. The defects that in the prior art, consumed time is long, and the hierarchical structure of a graph cannot be captured are overcome.

The invention provides a power supply network simulation method based on spectrogram rarefaction, and the method comprises the steps: building a weighted undirected graph corresponding to an SPICE netlist of a power supply network, a right-end item and a netlist Laplacian matrix corresponding to the weighted undirected graph through the SPICE netlist, and obtaining a sparse subgraph of the weighted undirected graph; establishing a sparse Laplacian matrix corresponding to the sparse subgraph, and removing rows and columns corresponding to the grounding points in the netlist Laplacian matrix and the sparse Laplacian matrix; carrying out Cholesky decomposition on the sparse Laplacian matrix LP to obtain a triangular matrix; setting a convergence threshold, taking the triangular matrix as a precondition sub-to operate a precondition conjugate gradient method to solve a linear equation set LGx = b, obtaining an approximate solution x, i.e., obtaining a simulation result of the power supply network. By the adoption of the scheme, the overall operation time of an iteration solution is accelerated by more than four times while the similarity with an original image is guaranteed, and the time for simulating the power supply network in the chip is greatly shortened.

A method and apparatus for analyzing cyber threat intelligence data. The method includes: acquiring first and second CTI graphs including first and second CTI data, respectively, classified based on a first classification item; classifying the first CTI data and the second CTI data based on a second classification item determined depending on the first classification item; outputting a graph similarity of the first and second CTI graphs determined based on a first CTI similarity between the first and second CTI data when the first and second CTI data belong to the same classification as a result of the classification; setting the first CTI graph and the second CTI graph to be included in one group when the graph similarity is equal to or greater than a threshold value; and outputting CTI information including the first and second CTI data for each group.