55 results about "Score matrix" patented technology

The invention provides a vehicle fault root cause diagnosis method, equipment, equipment and a medium, and the method comprises the steps: obtaining a vehicle target fault phenomenon, determining a to-be-detected fault event set based on a preset fault logic relation, constructing an initial scoring matrix, enabling a row vector to correspond to a fault event, enabling a column vector to correspond to the state feature parameters of a plurality of evaluation dimensions, and carrying out the detection of the fault event set; executing the detection task of the highest comprehensive score fault event, obtaining feedback data, updating the initial score matrix parameters based on the feedback data to obtain an updated score matrix, and calculating the comprehensive score of each event based on the updated score matrix again. Iteratively executing the detection task corresponding to the updated highest score event to identify whether the detection event is a fault root cause or not until the fault root cause of the target fault phenomenon is determined; according to the method, through a dynamic priority scheduling mechanism, real-time feedback data is fused in multi-dimensional evaluation, a high-value diagnosis task is executed preferentially, the resource consumption of traditional traversal diagnosis is remarkably reduced, and the fault positioning efficiency is effectively improved.

The invention relates to the technical field of tiny fault detection, and discloses an intelligent tiny fault detection method and system based on empowerment divergence, and the detection method comprises the steps: collecting offline production data at each moment in an industrial operation process; constructing a principal component analysis model and extracting a principal component; calculating a data score matrix; performing JS divergence test under a sliding window strategy; performing local outlier abnormal value detection on the sequence of the JS divergence statistics through a local outlier factor weighting method, and introducing an obtained local outlier weight factor into the JS divergence statistics; fitting data distribution through a smooth kernel function; calculating the weighted JS divergence statistic of the online production data, and judging whether a fault occurs or not; according to the method, JS divergence calculation of local outlier factor empowerment is introduced, and a principal component analysis technology is combined, so that an efficient and stable tiny fault detection model is constructed, and the detection capability of tiny faults under specific conditions or in a local area is improved.

The invention relates to the technical field of natural language processing, in particular to an entity relation joint extraction method and system based on deep learning. The method comprises the following steps: S1, preprocessing an input text; s2, converting the preprocessed word segmentation result into vector representation; s3, enhancing the understanding of a model on the context by adopting a BiLSTM, and introducing a self-attention mechanism to enhance the text feature modeling capability; s4, dynamically constructing a score matrix of a triple for relation prediction; and S5, outputting an identified triple result. The entity relationship joint extraction method and system based on deep learning are used for solving the problem of accurate capture of entity overlapping and semantic dependency relationship in a complex context, can automatically identify triads from structured and unstructured input texts, and output a triad identification result.

The embodiment of the invention relates to the field of target detection, in particular to a multi-modal detection model training method and device based on small sample prompt fine tuning. One embodiment of the method comprises the following steps: acquiring a training sample set and an initial multi-modal detection model; executing a training step: obtaining a text semantic embedding vector sequence, and splicing the first-layer text prompt vector sequence and the text semantic embedding vector sequence to obtain an initial text embedding vector sequence; an image block embedding vector sequence group is obtained, the first-layer image prompt vector sequence and the image block embedding vector sequence are spliced, and an initial image embedding vector sequence group is obtained; obtaining an image text similarity matrix group; performing fusion processing on each image text similarity matrix to obtain an image text alignment score matrix; generating a classification annotation image; determining a multi-modal detection loss value; and determining a multi-modal detection model. According to the embodiment, the complexity of a multi-modal detection model can be reduced, and the occupation of computing resources is reduced.

The invention relates to a natural resource asset management and territorial space planning system fusion method and system. The method comprises the following steps: encoding territorial space planning vector data and natural resource asset registration data to generate a space-time binding code; constructing a spatial topological graph and a time sequence diagram based on the codes, respectively extracting planning constraint feature vectors and asset dynamic feature vectors, and fusing the vectors to generate a planning asset coupling scoring matrix; recognizing a conflict unit set based on the planning asset coupling scoring matrix; based on the conflict unit set and a preset business rule base, performing classification processing on each conflict unit in the conflict unit set to obtain a conflict type; and based on the conflict type, generating a service execution instruction set through a preset reinforcement learning decision model. According to the method, through space-time unified coding and feature fusion processing, the accuracy of data association and the scientificity of decision making are improved, so that collaborative governance of territorial space planning and natural resource asset management is realized.

The invention belongs to the field of educational informatization, and discloses a student-teacher-student double-selection recommendation method and system oriented to individual requirements and used for assisting in lesson of graduate students. Calculating the demand similarity of both teachers and students, and screening part of candidate lists; establishing a scoring matrix according to the matched historical data; the singular value decomposition scoring matrix is a product of three matrixes, namely a student feature matrix, a teacher feature matrix and a diagonal matrix; optimizing a student feature matrix and a teacher feature matrix; calculating a prediction score matrix according to the optimized student feature matrix and the optimized teacher feature matrix; and for each user, obtaining a teacher or a student with the highest predicted score from the objects in part of the candidate list through the predicted score matrix, and selecting the teachers with the top M predicted scores or the students with the top M predicted scores as a recommendation list. According to the method, the teacher-student matching accuracy can be improved, similarity calculation can be carried out by utilizing attribute information of newly added students or teachers, and content-based recommendation is carried out.

The invention discloses a training method of an open set target detection model, a target detection method, equipment and a medium. The method comprises the following steps: acquiring decoder query of an input image and a target boundary prediction result; obtaining text embedding characteristics of the text information and an alignment score matrix between the text information and decoder query; determining a vision-language prompt according to the decoder query, and generating a target name prediction result based on the vision-language prompt; and constructing a joint loss item, a mask alignment loss item and a distillation loss item, and training based on the joint loss item, the mask alignment loss item and the distillation loss item. According to the invention, while the dependence on a large-scale data set is reduced, the training convergence is accelerated, the training efficiency is improved, and the detection performance is improved, so that the model performance of the open-set target detection model in pedestrian detection analysis and detection of any category of targets is effectively improved.

The invention belongs to the field of financial text sentiment analysis, and particularly relates to a financial entity and sentiment joint extraction method. In order to identify entities and emotional polarities thereof, the method comprises the following steps: firstly, carrying out initial representation on sentences by utilizing a pre-training model BERT-wwm-ext, then, obtaining a plurality of score matrixes of the sentences through a multi-head self-attention mechanism, and respectively constructing a graph convolutional network (GCN) for the plurality of score matrixes of the entities so as to obtain deep semantic representation of the entities, so that the deep semantic representation of the entities is obtained on the basis of the deep semantic representation. According to the method, a multi-head attention mechanism with entities as queries is utilized, interaction information representation between the entities and sentences is obtained, sentence information representation fusing the entities is established, and joint extraction of financial entities and emotional polarities of the financial entities is achieved through a Transform decoding layer.

The invention relates to an intelligent deep questioning method and system, and belongs to the technical field of natural language processing and intelligent evaluation. The method comprises the following steps: acquiring a candidate response, converting the candidate response into a standardized text, analyzing the text in combination with multi-dimensional monitoring data, positioning competency missing elements, generating a missing identification set, and marking a sorting question-chasing point; constructing a dynamic self-adaptive score matrix in which competency is linked with a questioning point, fusing monitoring data to calibrate scores and confidence, and generating a questioning technology intervention scheme; a virtual interviewer role library is preset, roles are dynamically matched, and the intervention logic and the missing identification set are combined to generate specialized questioning content; and when the core element capability element characterization is insufficient or the confidence coefficient does not reach the standard, generating a target element cognition questioning problem, and after element extraction and fusion, outputting a multi-dimensional competency score and a confidence coefficient report. According to the invention, the upgrading of the questioning from mechanical questioning to intelligent technical intervention is realized, the questioning targeting and the evaluation accuracy are improved, and reliable support is provided for accurate talent selection.

The invention discloses a geometric perception-based feature matching method, which comprises the following steps of: extracting key points, key line segments and plane point curvatures from images as real geometric features by using a constructor, generating semantic feature descriptors by using SuperPoint, applying a filter, and combining the features to construct an image for each image; the information transmitter is used for integrating information between all local and global features and neglecting unnecessary representations in unrelated foreground or background areas, and the output of the information transmitter is composed of updated feature vectors; and the matcher is used for screening matching pairs by adopting optimal transmission in SuperGlue in order to generate a final matching matrix, and generating a final matching score matrix by utilizing two independent softmax outputs of LoFTR. According to the feature matching method based on geometric perception provided by the invention, by combining geometric information and a learning scheme, the geometric consistency of features is enhanced, the matching precision is improved, the effect of local feature matching is enhanced, and the performance of camera pose estimation is improved.

The invention discloses a depth sensor anomaly detection method and system, a storage medium and electronic equipment. The method comprises the steps that an absolute depth value collected by a depth sensor and a relative depth value output by a monocular depth estimation model are acquired; global sorting is carried out on the absolute depth value and the relative depth value, a sorting difference matrix between the absolute depth value and the relative depth value is further calculated, and a sorting inconsistency score vector is obtained; obtaining a first-stage detection threshold value by adopting a fused adaptive threshold value strategy based on the sorting inconsistency score vector, and forming an initial anomaly set; for each point in the non-abnormal point set, constructing a comprehensive similarity scoring matrix after fusion; and accumulating the comprehensive similarity scoring matrix according to rows to obtain a second-stage scoring vector, and determining a final refined abnormal point set. According to the invention, by introducing the relative depth information generated based on the monocular image, the robustness and environmental adaptability of the depth sensing system can be significantly improved.

The invention relates to the technical field of named entity recognition, and discloses a text named entity recognition model and method. Wherein the model comprises an embedded layer, a multi-layer perceptron, a global context enhancement module, a word pair relative position coding attention module, a layered feature fusion module and a prediction layer; the embedding layer is used for generating a text embedding representation according to the text data; the multi-layer perceptron is used for converting the text embedded representation into a two-dimensional word pair matrix; the global context enhancement module is used for generating global context enhancement representation according to the two-dimensional word pair matrix; the word pair relative position coding attention module is used for generating word pair relative position coding attention representation according to the two-dimensional word pair matrix; the hierarchical feature fusion module is used for generating a score matrix according to the global context enhancement representation, the word pair relative position coding attention representation and the two-dimensional word pair matrix; and the prediction layer is used for outputting a named entity prediction result according to the score matrix. According to the invention, the accuracy of text named entity recognition is improved.

The present application discloses a medical text detection method, a training method of a model, and related devices. The method includes: obtaining a medical text to be detected; extracting semantic representations of each character in the medical text to be detected; predicting a plurality of score matrices based on the semantic representations of each character; and decoding a target medical entity and a target medical relationship in the medical text to be detected based on the plurality of score matrices. The above solution can simultaneously solve the problems of entity nesting and entity relationship overlap in medical texts.

The present invention proposes a feature matching method based on the idea of iterative matching, belonging to the field of visual image processing. Specifically, first, two original images to be matched are subjected to feature extraction to obtain feature points and descriptors. Then, the matching scores between the feature points are calculated using the descriptors to construct a similarity score matrix, and the Sinkhorn algorithm is used to optimize and obtain a matching relationship distribution matrix. Next, the matching relationship probability distribution is obtained from the matching relationship distribution matrix, and through the NMS non-maximum suppression method, groups of non-adjacent matching feature points are selected and their probabilities are marked as 0. Finally, it is judged whether the number of matching feature point pairs meets the requirements. If so, the matching result is output; otherwise, new descriptors are regenerated for each feature point, and the similarity score matrix is reconstructed. By reducing the non-maximum suppression radius, new matching feature points are obtained. The present invention greatly reduces the computational overhead of the attention mechanism while ensuring that the matching quality is not affected.

The present invention discloses a line segment matching method, comprising: obtaining a first line segment set of a current image and a second line segment set of a reference image; constructing a first neighborhood set; constructing a second neighborhood set; selecting a line segment to be matched from the first line segment set, and using the first neighborhood to obtain the neighborhood of the line segment to be matched; obtaining a first candidate line segment set of the line segment to be matched according to the neighborhood of the line segment to be matched and the second neighborhood set; performing significant matching or matching based on affine similarity according to the category of the line segment to be matched using the first candidate line segment set to obtain a candidate line segment set of the line segment to be matched; obtaining a candidate line segment set of each line segment in the first line segment set; constructing a score matrix using the candidate line segment set of each line segment in the first line segment set; and obtaining matching line segments of the first line segment set and the second line segment set by calculating the score matrix. The present invention also discloses a line segment matching system.

The present invention relates to a method for pattern recognition and optimization of a refining process based on big data, comprising the following steps: preprocessing the historical data collected during the refining process to obtain standardized data; processing the standardized data by using the principal component analysis method, establishing a model by using the score matrix and drawing a confidence ellipse; calculating the score matrix for the newly collected real-time samples, and projecting them onto the confidence ellipse; the samples projected inside the ellipse are normal points, which can be added to the historical data to establish a new model to achieve adaptive update of the model; the samples projected outside the ellipse are abnormal points, and fault tracing can be carried out according to the fault contribution rate; further, according to the original variables corresponding to the points in the confidence ellipse, an improved path optimization algorithm can be adopted to solve the adjustment method and path of the operating variables.

This application relates to a life cycle impact assessment calculation method suitable for GIS-LCA, including the following steps: S01) Obtaining the process spatial location set S p S02) Obtain a list D suitable for GIS-LCA; S03) Obtain the set of spatial locations of characteristic influencing factors S c S04) Construct the spatial data matrix C of the characteristic influence factors; S05) Construct the process location set P of the characteristic influence factors. s ;S06) will P s The parameterization process is transformed into a characteristic influence factor location matrix P; S07) Calculate the score matrix H of the characteristic influence factors in spatial location; S08) Based on the score matrix H, perform environmental assessment and interpretation of the target product. This application changes the existing LCIA calculation method that mechanically adopts the flow name correspondence method. This application can utilize geographic information system technology to achieve automatic flow correspondence by means of spatial relationships, so as to meet the GIS-LCA calculation requirements for LCIA.

The invention discloses a real-time interactive learning method and system for recording and broadcasting course transformation based on AI guided questions and answers, and relates to the technical field of wisdom classs.A knowledge point confusion scoring matrix M is constructed based on preset knowledge points, corresponding interactive instructions are generated for the preset knowledge points, and AI guided questions and answers are started according to the interactive instructions. Analyzing text character strings of students, updating a knowledge point confusion scoring matrix M according to an analysis result, taking the knowledge point confusion scoring matrix M as input, performing data fusion on the knowledge point confusion scoring matrix M and a JSON-format file of a preset scene template, outputting prompt word character strings of an AI role system and the JSON-format file of a finite-state machine, constructing an AI dialogue engine, and performing interactive circulation. The data tuple of the interaction round is stored as an interaction log sequence, a three-dimensional performance evaluation vector is generated, the three-dimensional performance evaluation vector is matched with a preset logic set, an action instruction is generated, and the problems of one-way infusion, knowledge disjunction and difficult application of recording and broadcasting classes are solved.

The application provides a high-efficiency fuzzy test method based on a score matrix, only normal basic blocks with a block score greater than 0 are subjected to a plug-in operation, and branches with a reaching probability of 0 are deleted through a pruning mode, which can prevent the fuzzy test of the application from generating new irrelevant test case inputs; therefore, it can be seen that the application avoids the fuzzy test tool of the application from exploring paths to irrelevant codes through the optimization mode; meanwhile, test cases in the seed pool of the application can trigger new paths reaching the target basic block or trigger vulnerabilities in the code to be tested, and the application selects a test case according to the final score and the mutation probability corresponding to each test case in the seed pool to perform mutation, and continues to test the vulnerabilities of the code to be tested according to the test case after mutation, so that the test case can reach the target basic block as much as possible through a high-score path, and the vulnerability mining efficiency of the software system to be tested is improved.

The invention discloses a character string feature extraction method and device, a program product and electronic equipment, and relates to the technical field of data processing.The method comprises the steps that word segmentation is conducted on a character string according to a first rule to obtain L words, then the L words are grouped according to M context types corresponding to the character string to obtain M key grouping sets, and then the M key grouping sets are grouped according to M context types corresponding to the character string; a score matrix is generated on the basis of N programming languages and M key group sets, each score is used for representing the probability that a character string determined according to each key group set belongs to the corresponding programming language, a target programming language and a target context type are determined according to the maximum score in the score matrix, and finally the target programming language and the target context type are determined according to the maximum score in the score matrix. And performing feature extraction on the character string according to the target programming language and the target context type. According to the method and the device, the technical problem of low system security caused by feature extraction of character strings written by an interpreted programming language in the prior art is solved.

The application discloses a kind of scene matching methods based on deep learning, first construct scene matching model, select 1 photograph and 4 different angle template images;Then the selected image is simultaneously input to full convolution twin neural network model;Next using the two branches of full convolution twin neural network model respectively extracts feature, generates five feature maps;Afterwards, the feature map is convolved, the similarity is measured, and a score matrix is generated;Then retrieve the maximum value coordinates in score matrix, obtain the position of matching target;Again, the matching result is compared, and the template image with the best matching effect is output as the matching result;Finally, the scene matching model is deployed to the unmanned aerial vehicle platform, and the visual pose estimation of the scene matching result is performed to realize the cross-view visual positioning of the unmanned aerial vehicle. The application improves the accuracy of unmanned aerial vehicle positioning, and lays a foundation for subsequent construction of a scene matching model with higher model accuracy and faster matching speed.

The application discloses a digital person video detection method based on facial semantic features, acquires real person voice visual video to constitute a training data set; constructs a digital person video detection model based on facial reality mode representation, respectively extracts facial reality mode representation features, mouth features and voice features, after multi-modal feature fusion, generates a matching score matrix according to the fused features; trains the digital person video detection model by using the training data set, generates a matching score matrix by using the trained digital person video detection model, and aggregates the matching score matrix to obtain the overall matching score of the to-be-detected video, so that the forgery detection is realized. The application proposes a unified audio and video forgery detection framework based on the facial reality mode, and improves the deep forgery detection performance in different scenes.

The invention provides a gene sequence alignment method and electronic equipment, the alignment process is divided into a seed search stage and a seed expansion stage, in the seed search stage, a plurality of seeds are determined for a read sequence to be aligned; acquiring a plurality of query sequences to be compared and corresponding reference sequences thereof based on the seeds in a seed expansion stage; expanding the length of each sequence by an integral multiple of the data width of the SIMD instruction, and initializing a score matrix; in the calculation process of each round of comparison score, the calculation range of the current round is determined in each score matrix, and then the comparison score is calculated by using an SIMD comparison instruction in the reverse diagonal direction of the score matrix; and after all rounds of calculation are completed, determining an optimal matching sequence according to the maximum comparison score. According to the scheme, the sequence length difference in the comparison process is effectively considered, a large amount of invalid calculation is reduced, the operation speed of the seed expansion stage is increased, and then the sequence comparison efficiency is improved.

The invention belongs to but is not limited to the technical field of recommendation systems, and particularly relates to a recommendation system hybrid filling optimization method and system fusing user information and historical score data, and the method comprises the steps: S1, carrying out the feature extraction and coding of the personalized information of a user, and dividing the user into a plurality of clusters with similar features through K-means; s2, performing statistical analysis on scoring behaviors in each cluster, calculating an average score of articles in the cluster, and calculating a historical average score of each user at the same time; s3, balancing the historical average score of the user and the cluster average score through a regularization strategy, filling missing values in the score matrix, and generating a more complete score matrix; calculating the similarity by using the filled score matrix, selecting K users with the highest similarity as neighbors according to a calculation result, predicting scores of the target users on non-scored items by using scores of the neighbor users on the target items, and selecting N items with the highest scores as recommendations for the target users.

The invention discloses a self-adaptive multi-source credit risk model fusion method and device, equipment and a medium, and relates to the field of artificial intelligence, and the method comprises the steps: selecting a sub-score with the optimal distinction degree as a main score, and carrying out the discretization binning processing of the main score; mapping the discretized main score level into an embedded vector to form an embedded matrix; inputting the embedded matrix into a preset neural network gate controller to obtain a gate control probability vector representing the activation probability of each sub-score; performing sparse processing on the gating probability vector to generate a sparse mask; splicing the main score and all the sub-scores into a score matrix, and performing adaptive selection on the score matrix by using a sparse mask to obtain a mask score matrix; and inputting the mask scoring matrix into a pre-trained neural network classifier to obtain a risk probability prediction result. According to the method, the model discrimination performance and the data use efficiency can be considered at the same time, so that the technical bottleneck faced by a traditional static linear fusion mode is broken through.

The application relates to a hypergraph-based dense subgraph detection method and device and a terminal device, which obtains feature sequences of multiple users, obtains an abnormal risk score matrix according to the feature sequences of the users, calculates an overall risk score of the abnormal risk score matrix, obtains an initial overall risk score, updates the abnormal risk score matrix, calculates an overall risk score of the updated abnormal risk score matrix, and if the new overall risk score is greater than the initial overall risk score, calculates the sum of each row of the abnormal risk score matrix, and determines that a user with a sum of 0 is a risk user, so that a risk object can be accurately obtained, misjudgment is avoided, and the recognition precision is high. Moreover, the calculation speed is fast, a greedy mode is adopted in the calculation of the dense subgraph, linear time complexity is met, the calculation has an advantage on a large amount of data, can be self-adaptive, does not require any prior knowledge and expert experience, and the scheme can automatically perform screening.

The invention relates to the technical field of natural language processing, and discloses an AI role-oriented multi-dimensional consistency test system, which comprises the following steps of: extracting core role characteristics of an AI role according to an application scene of the AI role so as to construct a dynamic role map of the AI role; quantifying the dimension weight of the role dimension corresponding to the dynamic role map; a basic test case library of the role dimension is established, and the basic test case library comprises a standardized problem set and a scene script, so that a dynamic derivative case of the role dimension is generated; a cross-dimension conflict use case of a dynamic derivative use case is introduced, a multi-layer test engine of role dimensions is established, and the multi-layer test engine comprises a surface language analysis layer, a logical reasoning verification layer and a value alignment evaluation layer; and calculating cognitive load indexes of the AI roles to establish a multi-dimensional scoring matrix of the AI roles, and calculating global consistency indexes of the AI roles through the multi-dimensional scoring matrix. According to the invention, the accuracy of the AI role multi-dimensional consistency test can be improved.

The invention discloses a speech recognition system based on a convolutional enhanced self-attention model of an improved scoring matrix. In the invention, a new relation calculation module (RCM) is introduced, so that the attention recovery decoding method is optimized. The module aims to mitigate error accumulation in an attention recovery decoding method, thereby reducing the influence of CTC decoding errors on subsequent attention recovery. In addition, the system also introduces a jump fusion module (SFM) to integrate shallow and deep features. The limitation that a Conformer encoder lacks inter-layer skipping connection is solved, and the capability of effectively capturing and utilizing contextual information of the model is enhanced. Comparison experiments show that a graph-based method improves the recognition rate, especially on a relatively small Uygur language data set. Compared with a baseline model, the method has the advantages that the character error rates are respectively reduced by 0.03%, 0.35% and 0.44%. On a GeneralSpeech16.1 Uygur language data set, the character error rate of the method is 4.19%, and is reduced by 1.56% compared with that of a baseline model.

The application relates to the technical field of geological prospecting, and provides a method for comprehensively optimizing a prospecting target area. In the method, a grade interval boundary limit value of a target area metallogenic possibility is determined, a comment set score model corresponding to the grade interval is constructed, a score matrix of a geochemical exploration index of the prospecting target area and a score matrix of a geological index are generated; then, according to a corresponding weight matrix, a membership score matrix of the geochemical exploration index and the geological index of the prospecting target area in the grade interval is calculated respectively, so as to determine an index score comment of the prospecting target area in the grade interval; and according to the grade membership of the prospecting target area in the grade interval and the interval upper limit value of the grade interval in the index score comment, the comprehensive membership of the prospecting target area is calculated, and the optimized prospecting target area is output by sorting the comprehensive memberships of all the prospecting target areas. Therefore, the optimized prospecting target area is scientifically determined through theoretical calculation, and the deviation and interference of human factors are effectively avoided.

The application discloses a movie recommendation method based on triple auto-encoder combined with a knowledge graph, which comprises the following steps: 1) encoding the comment information between the user and the movie into an emotional classification as the input of an automatic encoder; 2) merging the rating of the movie, auxiliary information and generated comment representation into a semi-automatic encoder for reconstructing the output, learning the low-dimensional feature representation of the extended information through the semi-automatic encoder, fusing the obtained low-dimensional feature representation into the original feature space of the movie, and inputting the new feature as additional information into the semi-automatic encoder model; 3) designing a serial connection of the semi-automatic encoder and the automatic encoder, and comparing the output prediction score matrix of the third auto-encoder with the original score matrix. The application can expand the features of the movie information by using the interaction information between the user and the movie and the knowledge graph, and process the expanded features through the auto-encoder, so that the more accurate recommendation for the user is achieved.