25 results about "Similarity computing" patented technology

The invention discloses a medical image treatment method and system for multi-source heterogeneous data, and relates to the field of medical image treatment, and the method comprises the steps: firstly, respectively extracting an image embedding vector and a text embedding vector from original medical image data and a description text thereof through a deep learning model; then, the vectors of the two different modes are fused, and a unified semantic embedding vector is formed; based on the unified vector, through semantic similarity calculation with a standard term library, candidate mapping can be automatically generated, and dependence on rigid artificial rules is eliminated. More importantly, a closed-loop mechanism of manual auditing-feedback learning is introduced in the scheme, high-confidence mapping is adopted automatically, low-confidence mapping is audited by experts, and an auditing result is absorbed into a mapping knowledge base, so that the system has continuous learning and self-evolution capabilities, and a semantic gap of cross-mechanism data can be eliminated more intelligently and more accurately.

The application discloses a business code processing method and device, a storage medium and an electronic equipment. It relates to the field of financial technology, and the method comprises the following steps: in the case that a business code repository is detected to have a business code modification event, a target code repository is matched according to a code file name corresponding to the business code modification event; in the case that a code class identical to the code file name is not matched, an attribute tree of a first code class and attribute trees of a plurality of second code classes are respectively constructed; similarity between the first code class and each second code class is calculated according to the attribute tree of the first code class and the attribute trees of the plurality of second code classes, and a target code class is determined from the plurality of second code classes according to the similarity. Through the application, the problem that the processing accuracy of business code is low when the code repository is adaptively modified in the related art is solved.

The application relates to an industrial process state monitoring method based on robust space-time joint projection, which comprises the following steps: data preprocessing and standardization; data space manifold structure and time dynamic information capturing: calculating the similarity degree between samples according to neighborhood overlapping order similarity, obtaining a data similarity matrix, and thus obtaining the overall space manifold information of process data; subsequently, calculating a data time derivative matrix, extracting slow change characteristics in the data, and obtaining key dynamic information; feature extraction and monitoring model building based on robust space-time joint projection: comprehensively integrating space-time information, performing dimension reduction projection on the data, obtaining key low-dimensional characteristics for retaining the representation of system operation states, constructing monitoring statistics and monitoring thresholds based on the key low-dimensional characteristics, and building an offline process monitoring model; real-time online monitoring of the state of an industrial system.

Techniques are disclosed that relate to skip lists. A computer system maintains a skip list having towers of varying depths and entries storing pointers to other towers. A first tower includes an entry at a particular depth storing a pointer to access an entry of a second tower. The pointer includes first similarity information indicating an amount of similarity between a key of the first tower and a key of the second tower. The computer system performs a traversal of the skip list for a search key. The computer system generates second similarity information indicating an amount of similarity between the first tower's key and the search key. Based on a comparison involving the first and second similarity information and without accessing the second tower to obtain information about its key, the computer system determines whether to traverse to the second tower using the pointer or descend the first tower.

The application discloses a content-tracing-based enterprise culture hot word analysis method, and relates to the technical field of enterprise culture analysis.The method comprises the following steps: collecting culture-related original data from three dimensions, and constructing an enterprise-specific culture term library through preprocessing, density peak clustering, de-redundancy and matching with top-level culture elements; classifying enterprise internal texts to be analyzed according to departments and text types, automatically matching terms of corresponding categories in the enterprise-specific culture term library, and performing directional word segmentation based on preset rules; calculating the scene coverage of terms according to department scenarios and text type scenarios, calculating the culture correlation degree by using Jaccard similarity, and screening core hot words by combining the Apriori algorithm; locking target hot word origin texts based on the core hot words, counting the track data of the origin text transmission channels, identifying the transmission nodes and constructing the transmission track; and the method improves the accuracy of hot word analysis and provides support for targeted promotion of enterprise culture.

This invention discloses a method and system for multi-source office information fusion processing based on big data, belonging to the field of office data processing technology. The method includes acquiring standardized multi-source office information that has been cleaned, standardized, and de-identified; performing feature extraction and quantitative weighting, initial cluster center optimization, hybrid distance calculation, and cluster iteration on the standardized multi-source office information to generate office information clustering results; combining the clustering results, performing multi-dimensional similarity calculation, temporal feature weighting, and two-layer association fusion to obtain office information association fusion results; classifying and integrating the association fusion results to generate and output fused office data. The system includes data preprocessing, big data fusion processing, and output display modules, enabling full-process fusion processing of office information, effectively breaking down information silos, simplifying office operations, and improving office efficiency.

This application discloses a face recognition method, apparatus, storage medium, and electronic device. The method includes: calculating the similarity between the facial features of a person to be identified and the facial features of candidate comparison persons to obtain a candidate comparison similarity; selecting candidate comparison persons from a pre-stored database; determining a candidate dynamic comparison threshold condition pre-associated with the candidate comparison persons; wherein the dynamic comparison threshold condition is determined based on the facial identifier of the comparison persons, the comparison similarity, and scene information; if the candidate comparison similarity meets the candidate dynamic comparison threshold condition, then the face recognition between the person to be identified and the candidate comparison persons is successful. This technical solution, based on an adaptive dynamic comparison threshold for face recognition, can effectively improve the scene adaptability and ease of use of face recognition.

This invention proposes a publicly audited Byzantine robust federated learning system and method. The invention introduces a third-party audit server to audit the federated learning training process. First, the aggregation server sends initial model parameters to each federated learning client and the audit server. Second, the federated learning client trains its model on its local dataset and uploads the local model update to the aggregation server. Then, the third-party audit server trains its model to obtain the audit server's model update and uses a random sampling method combined with cosine similarity to calculate the similarity between each client and the audit server. Finally, the aggregation server uses the similarity as a weight for each client and obtains the global model update through weighted averaging. This invention continues to execute the above process until the maximum number of training iterations is reached. This invention can still train an accurate model even when Byzantine clients are involved.

The application discloses an edge network malicious traffic detection method based on few-shot incremental learning, comprising the following steps: collecting network traffic data with labels, dividing bidirectional flow and selecting the first M data packets to construct a time sequence feature vector; taking a convolutional neural network as a backbone network to pre-train a feature extraction module; replacing a classifier with a prototype network, calculating sample adaptive weights by a cosine similarity to obtain a weighted prototype vector, and combining an adaptive margin CosFace loss function to optimize a model; and deploying the model to an edge environment to analyze traffic in real time and classify. The application retains flow dependency by bidirectional flow time sequence features, adapts to attack category incremental scenarios by few-shot learning, and enhances the fine-grained distinguishing ability of malicious traffic by an adaptive margin mechanism, so as to solve the detection difficulty of the edge environment sample scarcity and the attack type variability, and effectively improve the efficiency and accuracy of the edge network malicious traffic detection.

The invention relates to a carbon check domain knowledge graph construction method and system based on prompt engineering, and the method comprises the steps: obtaining a multi-source unstructured text including an enterprise carbon emission report, an industry check guide and a carbon accounting standard file, and carrying out the preprocessing, and obtaining a to-be-processed text fragment set; calculating the similarity between every two to-be-processed text fragments in the to-be-processed text fragment set to obtain an independent text fragment set; then semantic analysis is carried out, and theme keywords and corresponding weights are generated; carrying out keyword matching retrieval and semantic similarity calculation, and calculating comprehensive retrieval similarity; based on the comprehensive retrieval similarity, generating an enhanced cue word; based on the enhanced cue word, performing knowledge extraction on the text fragment to obtain a structured knowledge result and performing quality optimization; and based on the optimized knowledge result, constructing a carbon inspection domain knowledge graph. Compared with the prior art, the method has the advantage of high knowledge extraction precision.

The application provides an attack homologous analysis method, system, electronic equipment and storage medium, and relates to the technical field of network security. The method comprises the following steps: obtaining event clues, alarm data and sample data of a to-be-tested attack event; extracting ATT&CK tactics and technology features from the alarm data, extracting sample HASH values and code features from the sample data, and extracting attack industry and country features from the event clues; respectively calculating the similarity between the extracted features and corresponding features in a preset APT organization feature library; calculating a comprehensive similarity score based on each similarity, and determining the homologous APT organization of the to-be-tested attack event according to the comprehensive similarity score. The application integrates multiple-dimensional data sources, breaks through the analysis blind area caused by a single data source, simultaneously fuses static code and dynamic alarm features, effectively overcomes the false alarm problem caused by incomplete confusion code disassembly, and significantly improves the accuracy, comprehensiveness and robustness of attack tracing and homologous identification.

The application discloses a zero worker post intelligent matching and recommendation system based on a multi-dimensional dynamic label system, and relates to the technical field of zero worker post intelligent matching.The application first collects zero worker job-seeking and enterprise recruitment multi-source data through a label construction module, and constructs six-dimensional one-to-one correspondence zero worker dynamic labels and post dynamic labels;then a post matching module establishes a dimensional label group according to the dimensional correspondence relationship, calculates the similarity of each dimension by using the cosine similarity after weighting, and obtains the basic matching degree by weighting; subsequently, the comprehensive matching score is calculated by combining the supply and demand and cost factors, and the optimal combination is optimized according to the comprehensive matching score to complete the matching; the application solves the problems of traditional static, single-dimensional label description incompleteness and update lag from the label source, realizes the accurate description of the explicit and implicit attributes of zero workers and posts, greatly reduces the post mismatch rate, and significantly improves the matching accuracy and efficiency.

This application discloses a multimodal geological knowledge retrieval method, apparatus, equipment, medium, and product, relating to the field of geological data retrieval. The method includes: obtaining a user's query vector; calculating the similarity between the query vector and vectors in a multimodal geological knowledge vector database to obtain a similarity score; the vectors in the multimodal geological knowledge vector database being vectors in a unified semantic space; sorting the vectors in the multimodal geological knowledge vector database in descending order according to the similarity score to obtain retrieval results; and feeding back the retrieval results to the user and dynamically adjusting the vector weights in the multimodal geological knowledge vector database based on the user's feedback signal on the retrieval results. This application enables cross-modal and accurate retrieval.

The application relates to the field of automobile electronics, and provides a multilingual CAN bus analysis method and system based on dynamic resource mapping. The method carries out pretreatment on a multilingual term library and CAN original data to generate a standardized term dictionary and regularized time sequence data; based on the term dictionary and the time sequence data, a numerical correlation degree is calculated by fusing a Pearson correlation coefficient and mutual information, a semantic correlation degree is calculated by combining a term label similarity, multilingual terms and system context attributes are injected into a signal node, a comprehensive correlation degree score and an association type attribute are injected into an association edge, and a signal association network with enhanced semantics is constructed; interface multilingual real-time switching is realized through a dynamic translation dictionary, target language resources are loaded according to a language switching instruction, static interface text and dynamic data extracted from a network are real-time translated, and an interface is completely updated through a signal slot mechanism, so that problems such as insufficient semantic support, language switching needing restarting and incomplete interface updating are effectively solved.

The application discloses a kind of based on big language model's equipment purchase demonstration duplicate checking method, device and electronic equipment, belong to instrument and equipment management technical field, the equipment purchase demonstration duplicate checking method based on big language model includes: to be purchased equipment and inventory equipment are matched with name and category, then the equipment text description of the to-be-purchased equipment and the inventory equipment that satisfy name matching and category matching respectively are input into big language model for semantic analysis to obtain respective text vector, then the report text vector and the storage text vector are calculated to find out suspected duplicate equipment, and then generate duplicate checking analysis result;The application introduces multi-stage screening process and semantic similarity analysis mechanism, improves the identification ability of same or similar function equipment;By introducing candidate scale control mechanism, while ensuring the coverage of duplicate checking, the overall system computing overhead is reduced.

The application discloses a multi-objective vehicle path optimization method based on fuzzy time window satisfaction degree, which is suitable for a distribution scene containing a distribution center, vehicles and customer sites. The method first acquires customer coordinates, demand, time window and other data and calculates a distance matrix, generates a disturbed time window after disturbance on the basis of random disturbance of the deterministic time window, and represents the disturbed time window as a three-parameter fuzzy number [a, m, b]. The fuzzy arrival and service time are obtained through fuzzy operator recursion. Then, a quasi-trapezoidal membership function is constructed to calculate customer satisfaction and average satisfaction through barycenter and similarity. A heuristic and roulette are used to generate an initial population, and a distribution cost Cost and average satisfaction AvgMu are used as double targets. After non-dominated sorting, tournament selection, crossover and mutation evolution, the solution quality is optimized by combining local search, and the Pareto optimal solution set is output after the termination condition is met. The application integrates the uncertainty of the time window into optimization, balances the cost and satisfaction, and improves the path robustness.

This invention belongs to the field of data processing and discloses a multi-level interpretability method, system, terminal, and storage medium based on a large-scale multimodal medical vision model. The method includes: acquiring a target text report, performing entity recognition processing and attribution calculation to obtain a comprehensive attribution heatmap; performing structured decomposition processing on the target text report to obtain multiple smallest medical recognition units, and calculating similarity based on the multiple smallest medical recognition units and the comprehensive attribution heatmap to obtain an image-text matching score; inputting the target medical image and the target text report into a target result discriminator to obtain single evidence support and joint evidence support; and combining the comprehensive attribution heatmap, image-text matching score, single evidence support, and joint evidence support to obtain an interpretability result. This invention can effectively improve the interpretability and transparency of the model's output text report process, thereby enhancing the credibility of the text report output.

The application discloses a kind of BERT-based intelligent operation and maintenance alarm causal relationship analysis method.The method is upgraded by semantic modeling, joint coding device type and alarm content, so that the semantic similarity of core switch link interruption and access layer device off-network is improved by 40%, and the misjudgment rate is significantly reduced compared with independent coding scheme.Through the organic combination of deep semantic modeling and dynamic time sequence analysis, the intelligent operation and maintenance alarm causal relationship is realized from rule-driven to data intelligent-driven, which provides an efficient analysis tool for intelligent operation and maintenance of complex systems.The method realizes accurate causal modeling of complex system alarm data through semantic similarity calculation under time sequence constraint and multi-dimensional weight dynamic fusion, and is widely applicable to large-scale distributed system fault analysis and root cause positioning scenarios in industries such as power systems, communication networks, internet platforms, cloud computing infrastructures and industrial internet of things.

The application belongs to the technical field of Internet of Things security, and discloses a power system Internet of Things firmware multi-granularity vulnerability similarity detection method and device, wherein the method comprises the following steps: obtaining a to-be-detected firmware Hash value, and comparing the to-be-detected firmware Hash value with a vulnerability sample firmware Hash value; if the two Hash values are the same, it is considered that the to-be-detected firmware has a vulnerability; if the two Hash values are different, a file set Fs is obtained by extracting the to-be-detected firmware; the Hash value of a target file Ff in the file set Fs is compared with the Hash value of a vulnerability sample file; if the two Hash values are the same, it is considered that the to-be-detected firmware has a vulnerability; if the two Hash values are different, the feature attribute of the file Ff is extracted, and the similarity between the file Ff and the sample file is calculated according to the feature attribute; and according to the comparison between the similarity and a first threshold value, it is judged whether the to-be-detected firmware has a known vulnerability. The method can realize the vulnerability detection of firmware of power Internet of Things devices of multiple architectures, reduce the detection and analysis workload, and effectively improve the detection efficiency.

This invention relates to a method, system, equipment, and medium for technology transfer based on supply and demand matching. The method includes: acquiring basic data on scientific and technological achievements and user demand data, and constructing a feature set of scientific and technological achievements and a feature set of user demands; calculating the overall text feature similarity and key dimension feature similarity, and calculating the total similarity of key dimensions based on the key dimension feature similarity; calculating industry relevance based on industry attribute features; after obtaining technology maturity, industry transfer matching degree, and policy adaptability based on the basic data on scientific and technological achievements, calculating a rule correction coefficient; calculating the transfer matching degree after correcting the overall text feature similarity, total key dimension similarity, and industry relevance based on the rule correction coefficient, and generating a technology transfer recommendation strategy based on the transfer matching degree; the system, equipment, and medium are used to implement the above method. Compared with the prior art, this invention provides a method that can meet users' technology transfer needs.

Embodiments of the present application disclose a communication method, a communication device and a storage medium, which are applied to the technical field of communication and are used for adapting to dynamically changing node scale, resource distribution and diversified task characteristics, thereby reducing scheduling overhead and maintaining load balance. The method of the embodiments of the present application comprises: obtaining first information of a plurality of clusters, the first information being used for calculating task similarity between the plurality of clusters, the task similarity being used for indicating the similarity degree of task types between different clusters in the plurality of clusters; and sending second information, the second information being calculated according to the task similarity, the second information being used for indicating a first grouping, the first grouping comprising part of the clusters in the plurality of clusters. In the embodiments of the present application, a grouping strategy control function takes the task similarity of different clusters as an important basis for cluster grouping, thereby improving the probability of obtaining a dependent image in a target cluster when a task in a grouping is migrated across clusters and accelerating the processing of the task.