115 results about "Similarity model" patented technology

The invention belongs to the technical field of machine room safety monitoring, and discloses an image review method fusing semantic understanding and visual recognition, which comprises the following steps: calculating visual / semantic feature dynamic credibility in real time through an exponential weighted moving average algorithm in combination with environment interference and equipment state parameters; resNet50 is adopted to extract visual features in global and local branches, and a BERT model is adopted to encode semantic features; correcting the feature correlation degree according to the scene, calculating a dynamic weight, carrying out heterogeneous calibration through an attention mechanism, and calling priority rules such as'physical security features are higher than behavior features' for arbitration during conflicts; the method is advantaged in that low-credibility feature interference fusion is avoided, abnormity identification accuracy in a complex scene of a machine room is improved, multi-modal data cooperation demands are adapted, scene labels are marked based on time, work orders and historical data dimensions, an exclusive sub-model is constructed for a high-density scene through DBSCAN density clustering, and low-frequency scene parameters are migrated to a similar model.

A method for delineating an ecological corridor based on a maximum similarity model determines ecosystem type according to the determined ecological source, and assigns a habitat suitability index; and according to a maximum similarity matrix between different ecosystems, assigns the habitat suitability index to peripheral pixels of the selected ecological source, to obtain a similar value of the habitat suitability index. The habitat suitability index is corrected according to the surface curvature data and population density data, and resistance surface data is calculated according to the corrected value. Based on the resistance surface data, an ecological corridor is delineated comprehensively to find one or more paths corresponding to a minimum cost distance as ecological corridors. The influence of ecological environment on ecological corridor establishment is considered more comprehensively, so that the ecological corridor is delineated more accurately, and applied into practice to maximum extent.

The invention provides a large-model multi-label classification method, system and device based on ReAct and a vector library, and belongs to the technical field of artificial intelligence. The method comprises the following steps: constructing a multi-level label mapping tool according to each label data table corresponding to a preset multi-level label system; and constructing a label retrieval tool based on a pre-trained text similarity model and a preset label vector database. Constructing a ReAct tool chain based on a preset label dynamic adjustment tool, a multi-level label mapping tool and a label retrieval tool; the preset label dynamic adjustment tool is constructed on the basis of cue words and a large language model and is used for updating candidate labels according to the to-be-labeled text corpus, the candidate labels output by the label retrieval tool and label rules; and based on a preset ReAct cue word project, the large language model and the ReAct tool chain, constructing a large model multi-label classification module so as to input the to-be-labeled text corpus from the user terminal into the large model multi-label classification module for multi-label classification.

The invention discloses a multi-parameter distortion compensation method for a fractured rock mass seepage similar model test, and relates to the technical field of rock mass engineering physical model tests. The method comprises the following steps of: firstly, determining two constraint conditions that the fluid property is invariable and the gravitational acceleration is invariable, correspondingly selecting a basic physical quantity and extracting a core criterion; then analyzing three types of distortion sources of geometric dimensions, dielectric materials and boundary conditions, and establishing a distortion function according to result parameters of seepage flow velocity, seepage flow and seepage pressure; secondly, parameters difficult to control are selected as distortion items, parameters easy to regulate and control are selected as compensation items, the distortion items are substituted into a distortion function to obtain a distortion ratio, the optimal distortion ratio of the compensation items is solved based on a compensation balance formula, and the parameters are adjusted; and finally verifying the interval range of the actually measured distortion ratio. According to the method, the multi-source distortion coupling influence is accurately quantified, the compensation pertinence and precision are improved, operation is easy and convenient, universality is high, and reliable support is provided for underground engineering seepage research.

The invention discloses a rainfall image sequence similarity search method based on a twin neural network. The rainfall image sequence similarity search method comprises the steps of obtaining a target rainfall image sequence and a target rainfall image library; constructing a rainfall image similarity model; based on the target rainfall image sequence and the target rainfall image library, performing training optimization on the constructed rainfall image similarity model to obtain a target rainfall image similarity model; and based on the target rainfall image sequence and the feature vector of the corresponding frame in the target rainfall image sequence extracted by the target rainfall image similarity model, carrying out rainfall image sequence similarity search on the target rainfall image library to obtain a rainfall image sequence similar to the target rainfall image sequence. According to the method, deep learning and a meteorological mechanism are fused, efficient similarity search of a rainfall image sequence is realized by constructing a rainfall image sequence similarity search model of a feature extraction network and twin neural network architecture, and three bottlenecks of feature robustness, time sequence modeling and calculation efficiency are broken through.

A method implements pre-trained large language model driven bug localization. The method includes receiving a report and applying a fine-tuned language model to report text from the report, to source text from a source file of a set of source files, and to commit text from a commit of a set of commits to respectively generate a report vector, a source vector, and a commit vector from the fine-tuned language model. The method further includes applying a similarity model to the report vector and the source vector to generate a report source score and includes applying the similarity model to the report vector and the commit vector to generate a report commit score. The method further includes applying a ranking model to the report source score and the report commit score to identify the source file corresponding to the report and includes presenting the source file responsive to the report.

Systems, media, and computer-implemented methods are provided for identifying similar chunks of text to tune a text similarity model, such as a text similarity model that is used to find content in response to queries. Using a masked language model, a machine learning model may be tuned on different content from that which the machine learning model was trained. The machine learning model as tuned may be used to determine vector embeddings for terms in chunks of content. Chunks may be matched to each other by finding a term in one chunk having a highest similarity score with a corresponding term in another chunk. Aggregate similarity scores may be determined between the chunks based on the term-to-term similarity scores. If an aggregate similarity score for a pair of chunks satisfies one or more conditions, a text similarity model may be tuned to identify the pair as similar.

The invention provides a hybrid filtering denoising method and device based on adaptive parameters, and relates to the technical field of three-dimensional point cloud processing. The method comprises the following steps: acquiring three-dimensional point cloud data of a target object; performing voxel segmentation processing on the three-dimensional point cloud data to obtain a plurality of voxels; according to the segmented three-dimensional point cloud data, combining a plurality of voxels by using a density similarity model to obtain a voxel set with different density distributions; based on the merged voxels, a radius filtering parameter and a statistical filtering parameter of each voxel set are adaptively determined, the radius filtering parameter is used for filtering isolated noise in the three-dimensional point cloud data, and the statistical filtering parameter is used for filtering random noise in the three-dimensional point cloud data; performing random noise removal processing on the three-dimensional point cloud data according to the statistical filtering parameters; and performing isolated noise removal processing on the three-dimensional point cloud data after random noise removal according to the radius filtering parameter.

The invention discloses a tobacco marketing abnormal address management and control method, system and device based on GIS and multi-dimensional data fusion and a medium, and belongs to the technical field of tobacco industry informatization and internet marketing. The technical problems that in the prior art, address risk control dimensionality is single, query precision is insufficient, data linkage is lacked and visual analysis capacity is weak are solved. According to the technical scheme, the method comprises the steps that abnormal address data is constructed and queried based on a multi-dimensional label; gIS visualization and correlation analysis based on a mixed similarity model: GIS positioning and range loading, spatial candidate set filtering, text similarity fine screening, mixed weighted scoring and visualization rendering and statistics are realized through the mixed similarity model of spatial index coarse screening and text algorithm fine screening, and correlation order detail form display based on GIS range triggering is carried out.

The invention discloses a multi-stage attention and generative adversarial network collaborative product generation method and system, and relates to the technical field of image generation, and the method comprises the steps: obtaining a low-resolution clothing image, focusing the low-resolution clothing image on a clothing image key region through a multi-stage attention mechanism, and carrying out the image refining processing through combining text features, a high-resolution clothing image is obtained; processing the text description to obtain corresponding word features, processing the high-resolution clothing image obtained by processing the multi-stage attention generative adversarial network to obtain corresponding local image features, and inputting the word features and the local image features into a deep attention multi-modal similarity model to obtain a deep attention multi-modal similarity model; outputting and obtaining the fine-grained similarity loss of the image and the text; and optimizing a pre-constructed multi-stage attention generative adversarial network based on the adversarial loss and the fine-grained similarity loss of the image and the text, and generating a product image based on the optimized multi-stage attention generative adversarial network.

In a computer-implemented method and associated tangible non-transitory computer-readable medium, an image of a damaged vehicle may be analyzed to generate a repair estimate. A dataset populated with digital images of damaged vehicles and associated claim data may be used to train a deep learning neural network to learn damaged vehicle image characteristics that are predictive of claim data characteristics, and a predictive similarity model may be generated. Using the predictive similarity model, one or more similarity scores may be generated for a digital image of a newly damaged vehicle, indicating its similarity to one or more digital images of damaged vehicles with known damage level, repair time, and / or repair cost. A repair estimate may be generated for the newly damaged vehicle based on the claim data associated with images that are most similar to the image of the newly damaged vehicle.

A system and method for channel layout evaluation including: a computer processor and a channel detective service executing on the computer processor and including functionality to: receive a request to perform channel layout evaluation on a media item including a provided set of channels; perform metadata extraction on the media item to generate a metadata representation of the media item and to identify at least one channel layout discrepancy of the provided set of channels; perform layout detection using the metadata representation by executing a similarity model configured to generate a mix group comprising at least a subset of the provided set of channels, annotating a primary language of the mix group, and annotating a service type of the mix group; and updating the metadata representation with the annotated mix group, wherein the updated metadata representation is provided for streaming the media item.

The invention discloses a denture three-dimensional model feature construction and matching method based on curvature adaptive sampling, which comprises the following steps: S1, acquiring denture surface point cloud data through three-dimensional scanning, converting the data into an STL model, and comparing a historical STL model pre-stored in a folder; s2, constructing a self-adaptive sampling weight, selecting a triangular patch through replacement random sampling, and generating a sampling point; s3, carrying out normalization processing on the sampling points; s4, calculating the distribution of D2 and A3 by using the normalized sampling points, and obtaining corresponding shape features; s5, splicing the features to form a joint geometric feature vector; and S6, performing S1-S5 on the to-be-matched false tooth to obtain a matching vector, performing S2-S5 on the historical models one by one to obtain comparison vectors, calculating cosine similarity between the matching vector and each comparison vector, and outputting a Top-K similarity model. According to the method, key geometric structure feature expression is enhanced, robustness to point cloud density change, local defects and the like is high, and matching is ensured to be reliable and consistent.

The invention discloses a remote bank fuzzy question-answer accurate matching method based on deep learning. The method comprises the following steps: firstly, constructing a business semantic system (clear business class-scene-specific intention), a term and variant mapping library and a sentence pattern template containing a replaceable entity slot position; extending the original similar question pairs to generate derivative sentences and candidate similar sentences, filtering, clustering to generate cross-sentence pattern similar pairs, and selecting negative samples to control the literal overlapping rate; then constructing a twin BERT network, mixing original and enhanced samples into a training set, and training in two stages by using improved triple loss; finally, sentence vectors are generated through the trained twin network, semantic similarity is judged according to cosine similarity, question pairs which are judged to be similar through a model are subjected to secondary verification through a service rule base, and score reduction is carried out if the question pairs do not pass verification.

The invention provides a bridge deck reduced scale model design method and experiment system in a plateau environment, and the design method comprises the steps: building a multi-factor coupling similarity criterion based on a similarity principle, physical parameters affecting the prototype structure performance of a bridge deck, and plateau environment parameters; determining a dynamic stress similarity ratio model based on a multi-factor coupling similarity criterion and the plateau environment parameters; based on the dynamic stress similarity ratio model and the section shape and size of the bridge deck prototype structure, performing scale design on the bridge deck prototype structure to obtain an initial scale model; and performing simulation experiment and simulation verification on the initial scale model based on a preset numerical simulation strategy and a preset plateau environment simulation strategy, and performing parameter optimization of the dynamic stress ratio similar model to design a target scale model. According to the scheme provided by the invention, experimental simulation of the bridge deck in the plateau environment can be efficiently and accurately carried out, and the experimental simulation cost is reduced.

The invention discloses an image restoration method, which comprises the following steps: acquiring a target image, a reference image set and a boundary intensity image, and extracting target features, a reference feature list and boundary features; based on the extracted features, performing similarity modeling by using a learnable spiral path to generate a similarity model result; based on a similarity model result, performing geometric alignment and double-domain fusion on the reference feature list to generate initial fusion features and alignment parameters; calculating an error mode pattern based on the initial fusion features; executing closed-loop feedback according to the error mode pattern, and iteratively updating to generate a final fusion feature; and decoding the final fusion feature to obtain a repaired image. Through spiral sampling of boundary perception, non-folding alignment and closed-loop feedback, the structure consistency and detail fidelity of the repaired image under non-rigid deformation are improved.

The invention relates to the technical field of building information models and intelligent design, and provides a bridge steel structure collaborative design method and system based on the BIM technology, and the method comprises the steps: obtaining a plurality of bridge steel structure BIM models from a BIM database, and design parameter vectors, structure topological graphs and collaborative design data vectors thereof; constructing a structural feature extraction model by training a BIM structure encoder model, calculating the similarity between the models, and retrieving a similar model; establishing a BIM data fusion model, and training by combining the design parameter vector of the target model and the collaborative design data vector of the similar model to obtain a collaborative design prediction model; and searching a similar model for a to-be-collaborative design model, and generating collaborative design data through a prediction model based on design parameters of the to-be-collaborative design model and collaborative data of the similar model. According to the method, intelligent collaborative optimization of bridge steel structure design can be realized, the design efficiency and accuracy are improved, and manual intervention and repeated work are reduced.

The invention relates to the technical field of aero-engines, and discloses a dynamic tester design method and system for simulating a complex rotor system, and the method comprises the steps: constructing a three-dimensional entity similar model of a dual-rotor supporting system, and analyzing the modal shape vector of each order of a one-dimensional dynamic parameterized model of the dual-rotor supporting system. Calculating the parameter modal confidence coefficient and the associated modal confidence coefficient of the similar model according to the modal shape vector of the similar model and the modal shape vector of each order of the similar model, and adjusting the similar model by taking the two modal confidence coefficients as evaluation criteria to obtain a similar model with high confidence coefficient; according to the total rigidity, the supporting rigidity, the squirrel cage rigidity and the end seal rigidity of the similar model, an oil film damper rigidity reference value is obtained through analysis, the oil film damper structure is adjusted, and therefore the oil film damper structure meeting the design requirement is output. According to the invention, the bearing environment of a real aero-engine rotor can be simulated with high fidelity, and the problem of insufficient measurement dimension in a rotor synchronous monitoring scheme of a double-rotor tester is solved.

The application relates to a method for training a text similarity model, comprising: obtaining one or more original offer text pairs; performing the following operations on each of the one or more original offer text pairs: generating a first similarity score of the original offer text pair based on a first text similarity model; extracting an offer feature parameter corresponding to the original offer text pair; correcting the first similarity score based on the offer feature parameter to generate a second similarity score by using a preset correction formula; performing floating sampling on the offer feature parameter to generate a plurality of enhanced offer feature parameters, and restoring each of the plurality of enhanced offer feature parameters to an enhanced offer text pair; and performing model training by using an enhanced training set to generate a second text similarity model, wherein the enhanced training set comprises the one or more original offer text pairs and the enhanced offer text pair generated based on each original offer text pair.

Techniques for the use of verbal commands in human-robot communication. The number of tasks the robot can perform is limited to a specific set, while providing syntactic flexibility to users. The system includes two components: a speech recognizer for speech-to-text conversion and a natural language understanding module that maps the text to a command for the robot. After speech is transcribed to text, a nearest neighbor classifier can be applied in the high dimensional space of embedding tokens. Multiple variants of each command are provided in a database of reference embeddings, and the classifier can identify the k nearest reference embedding tokens to determine the command. The text similarity model allows for quick detection solutions to be deployed locally on a robot or other device. Local deployment reduces potential latency caused by a cloud connection, which can be important in many assistant robot applications.

The invention provides a method for constructing a bearing test distortion model of an ice-resistant structure under the action of a dynamic ice load, and belongs to the field of test similar reduced scale model design under an ice breaking load of the ice-resistant structure. The problem that in the prior art, the strain rate of a fully-similar model is dissimilar due to the scale effect is solved; a traditional distortion model design method has the problem of dissimilar energy distribution. The method comprises the following steps: establishing a fully similar model from an ice-resistant structure prototype to a geometric and plate thickness equal ratio reduced scale; establishing a plate thickness distortion reduced scale model by using the same material on the basis of the geometric similar reduced scale model; preliminarily determining a collision contact position of the thickness distortion reduced scale model; a dynamic ice load similarity criterion is applied, correction factors are deduced through dimensional analysis, and ice body parameters and collision speed of a distortion model system are adjusted, so that the ice breaking capacity and dynamic structure response are similar to those of a prototype system. The device is mainly used in the field of polar region icebreaking.

A URL is input into a virtual browser of a crawler which requests that URL and all of its resource URLs in order to form a network graph. Network features are extracted. An identity stage uses an identity model based upon lexical and host features to determine if the input URL and resource URLs are malicious or benign. A similarity stage uses a similarity model and similarity functions to produce a similarity score between each pair of the input URL and a resource URL. An activity stage uses an activity model to determine if the input URL's network is malicious or benign based upon the network graph and extracted network activity features. Edges of the graph are weights, similarity scores, attributes or network features. A finality stage concludes malicious if the identity stage concluded that the URL was malicious or the activity stage concluded that the URL's network was malicious.

The application discloses a kind of long approach in paste flow slope angle prediction method, belong to filling technical field, solve the prediction of filling slope angle in prior art is based on experience formula or static homogeneous model, unable to consider the influence of waste rock dynamic settlement on slurry flow resistance and slope form Problem, the method includes based on paste raw material basic parameter construction paste similarity model, based on range analysis combined with interaction analysis influencing factor on the influence degree of flow slope angle, convert flow slope angle calculation model into slope angle satisfaction function, obtain slope angle optimal filling parameter;The present application can comprehensively study the influence of multiple influencing factors and their interaction on paste flow slope angle through the pre-design and implementation of orthogonal experiment scheme, avoids the one-sidedness that single factor analysis may bring, makes the influence law obtained more comprehensive and in-depth, and is different from the homogeneous flow model or simplified discrete element model that the existing numerical simulation method is mostly used.

A method implements the use of large language models to generate bug localization explanations enhanced by code summarization. The method includes executing an explanation similarity model using a training report and a training explanation to generate an explanation score for an explanation sample. The method further includes filtering multiple explanation samples using multiple explanation scores to generate a set of filtered explanation samples. The method further includes executing a summarization similarity model using source code and a description to generate a summarization score for a summarization sample including the source code and the description. The method further includes filtering multiple summarization samples using multiple summarization scores to generate a set of filtered summarization samples. The method further includes training a language model using the set of filtered explanation samples and the set of filtered summarization samples to generate a fine-tuned model.

This invention discloses a method and system for identifying cross-connection issues in low-voltage distribution areas. It acquires voltage and current time-series data of users in the low-voltage distribution area, cleans the data using a mask interpolation strategy, and adaptively extracts voltage drop and current surge features based on the median absolute deviation algorithm. It constructs a voltage trend similarity model based on local regularization kernels and a mutation co-occurrence similarity model based on event level weights, employing a dual-track strategy combining daily rigorous assessment and feature revival to generate suspected voltage correlation pairs. A density-based clustering algorithm is used to generate an initial population, and a voltage similarity-weighted centroid screening mechanism is introduced to remove loose outliers in the feature space. A pairwise cross-validation model is constructed based on electrical causality, and the population is topologically refined using relative voltage response criteria and a maximum fully connected clique extraction algorithm to determine the final cross-connection population. This invention significantly improves the accuracy and robustness of cross-connection identification.

The invention provides a near-infrared displacement monitoring method and system based on multi-reference point correction, and the method comprises the steps: obtaining a near-infrared target image sequence, and extracting a candidate reference point set and a measuring point set; performing multi-dimensional feature extraction on the two to obtain a feature vector set; inputting the feature vectors into a preset multi-dimensional similarity model for similarity evaluation, and generating a comprehensive similarity matrix; self-adaptive matching is carried out based on the matrix and a similarity threshold value, and a reference point-measuring point matching relation table is established; and finally, performing differential correction on the measuring point set according to the matching relation to obtain a measuring point displacement value set. According to the near-infrared displacement monitoring method and system based on multi-reference point correction provided by the invention, the reference stability is improved through multi-reference point screening and multi-dimensional feature extraction, and adaptive matching is performed through the preset multi-dimensional similarity model to ensure the matching accuracy, so that the displacement data accurately reflects the real deformation of the structure, and the accuracy of the real deformation of the structure is improved. And the accuracy and robustness of a monitoring result are improved.

The invention relates to an engineering site residual deformation prediction method and device based on a physical similarity model. The method comprises the following steps: acquiring earth surface residual deformation time of a target monitoring area and observation time of a preset physical similarity model experiment; determining a time similarity ratio and a first geometric similarity ratio based on the earth surface residual deformation time and the observation time; obtaining prototype residual deformation of the target monitoring area and model residual deformation obtained by monitoring through monitoring equipment; determining a second geometric similarity ratio based on the prototype residual deformation and the model residual deformation; selecting a minimum value in the first geometric similarity ratio and the second geometric similarity ratio as a target geometric similarity ratio; and performing amplification processing on the residual deformation simulation value by using the target geometric similarity ratio to obtain a residual deformation prediction value, and performing amplification processing on the residual deformation simulation time by using the time similarity ratio to obtain a residual deformation duration prediction value. According to the scheme, long-term prediction of the earth surface residual deformation of the target monitoring area is realized.

The invention discloses an intelligent determination method and device of a network strategy and a storage medium, relates to the field of communication, is used for accurately matching the network strategy for network equipment, and comprises the following steps: receiving strategy request information; the strategy request information comprises resource demand information; determining at least one candidate network strategy from a strategy knowledge base based on the strategy request information and a similarity model; the strategy knowledge base comprises a plurality of network strategies, and the similarity between each candidate network strategy in the at least one candidate network strategy and the strategy request information is greater than a first threshold value; determining a policy score of each candidate network policy in the at least one candidate network policy based on the policy request information, the at least one candidate network policy and a reordering model; based on the strategy score of each candidate network strategy, determining a target network strategy meeting a matching degree requirement from the at least one candidate network strategy; the target network strategy comprises a strategy name and API information. The method and the device are applied to a network policy matching process.

The invention discloses an adverse reaction recognition and quick response system and method in an infusion process, and solves the problems that an individual feature similarity model and an individual-drug interaction prediction model in an existing method are static prediction models based on historical data, and dynamic changes cannot be captured in real time; the infusion monitoring method comprises the following steps: acquiring infusion monitoring information in real time; the adverse reaction recognition model recognizes and analyzes the abnormal monitoring data based on risk clustering analysis, calls an adverse reaction strategy in a response database, and triggers an adverse reaction pushing instruction; according to the system, effective infusion monitoring information can be obtained in real time in the infusion process, deep processing such as filtering and noise reduction and wave band periodical detection can be conducted on multi-source infusion monitoring information, dynamic changes in the infusion process are captured in real time through the adverse reaction recognition model, adverse reactions are found in time, and the accuracy of infusion monitoring is improved. And real-time monitoring and rapid response to adverse reactions are realized.

The present application relates to the field of artificial intelligence, in particular to a case retrieval method based on multiple models. Various multi-source data are collected and integrated; the semantic dependency relationship between the text fragments of the contradiction is captured by using Bert, and the similarity of the case data is calculated by using the BM25 algorithm; a model is trained by Bert according to twelve types of classifiers, and at the same time, a model is retrained for the overall similarity model training. The similarity model constructed by combining the case classification model in the prior art and the present application completes the accurate case retrieval of diversified and long-length contradiction dispute data. The legal case data of the present application is more specific and perfect, contains sufficient legal knowledge, can cope with the change of legal rules, the unification of diversified legal data, accurate collection, efficient auxiliary analysis of legal cases and resolution work. The present application is more accurate in calculating the similarity, and the relevant category of legal cases is recommended accurately.