84 results about "Similarity data" patented technology

The invention relates to the technical field of electrical digital data processing, and discloses a statistical entropy and multi-factor fused text similarity data processing method, which comprises the following steps that: a processor extracts substring sets which do not contain maximum common values of a first data sequence and a second data sequence, and calculates the quadratic sum of the lengths of substrings to generate local statistical entropy; traversing the maximum common substring set to obtain storage address indexes of the maximum common substring set in the first data sequence memory space and the second data sequence memory space, and constructing a topological mapping vector of a mapping structure displacement relationship; calculating the total number of inverted pairs of the topology mapping vector by using a merge sorting algorithm, and generating a normalized topology dissipation index; and by taking the local statistical entropy as an information carrier and taking the topological dissipation index as a structural damping factor, executing nonlinear damping modulation operation to obtain a final similarity score, and solving the technical problem that the block-level displacement cannot be identified by linear scanning logic by quantizing topological entropy increase of data distributed in a storage space.

The invention provides a sea area digital twinning method and system based on buoy and station data fusion, belongs to the related technical field of sea area digital twinning, and establishes a data quality evaluation mechanism and a similarity data compression mechanism by constructing a multi-source heterogeneous ocean data acquisition system. Intelligent fusion processing of buoy data and station data is achieved, and the problem that in the prior art, the data fusion precision is low is solved; through the self-organizing growth network structure of the ocean dynamic response enhancement model, neuron connection can be automatically adjusted according to the spatial-temporal variation degree of training data, an optimal topological structure is formed, and the defect that a traditional fixed parameter model cannot adaptively process a complex ocean environment is effectively overcome; by establishing a dynamic parameter adjustment mechanism and a multi-algorithm adaptive switching strategy, the most suitable prediction algorithm is automatically selected according to the correlation characteristics of the ocean data sequence, and high-precision digital reproduction of the complex ocean environment is realized.

The invention discloses a computing power resource scheduling method based on artificial intelligence, and particularly relates to the technical field of resource scheduling. According to the method, semantic similarity data and real-time computing power consumption data of tasks in the training and reasoning process of a multi-task artificial intelligence model are collected, an initial data mapping model is constructed, dynamic association features between task semantic features and real-time computing power demands are analyzed, computing power demand nonlinear fluctuation features driven by task semantics are recognized, and a multi-task artificial intelligence model is constructed. Generating a force demand mutation prediction result and a computing force demand dynamic prediction result; the computing power demand mutation prediction result and the computing power demand dynamic prediction result are combined, a dynamic scheduling strategy is optimized, computing power resource allocation in the multi-task artificial intelligence model training and reasoning process is adjusted in real time, local resource redundancy and global resource scheduling imbalance are eliminated, and the overall use efficiency of artificial intelligence computing power resources is improved.

The invention discloses a bank intelligent operation knowledge base implementation method and system based on a large language model and RAG, and the method comprises the steps: firstly extracting business terms from structured, semi-structured and unstructured multi-source heterogeneous multi-modal data of a bank, and then constructing a business dictionary containing term similarity data and an ontology tree; quantitatively calculating the release mechanism weight, the timeliness coefficient and the punishment association degree of the supervision clauses to obtain authority values, and sorting the authority values; then using a double-constraint loss function to finely adjust the pre-trained large language model as a vector representation model, vectorizing service data, and storing the vectorized service data into a vector database; and finally, aiming at user query, retrieving in a vector database, rearranging in combination with a supervision clause sorting result, and generating a business response after compliance verification. According to the method, the bank unstructured knowledge processing capability and retrieval precision can be improved, and efficient intelligent operation is realized.

The invention discloses a multi-view graph converter incremental learning method for identifying related diseases. The method comprises the steps of 1, heterogeneous network construction, 2, mask contrast learning, 3, graph converter feature extraction, 4, channel attention fusion, 5, depth matrix decomposition prediction and 6, incremental learning updating. According to the method, a heterogeneous network is constructed to integrate multi-source similarity data, mask contrast learning is utilized to generate multi-view embedding so as to reduce noise, node position information is coded by means of a graph converter, multi-view features are fused through a channel attention mechanism, disease association is further predicted through depth matrix decomposition, and the disease association prediction accuracy is improved. And an incremental learning strategy is designed to realize dynamic updating of new data, a new result is obtained under the condition that original input data and model retraining are not needed, and the method can adapt to the condition of explosive increase of the data.

The invention discloses a power supply control method and system based on a real-time slope adjustment algorithm, and the method comprises the steps: monitoring the load impedance characteristics of a power supply and environment noise data in real time, and obtaining an independent control time slice set; obtaining a time sequence identifier of the control time slice set, and obtaining a dynamic mode parameter packet bound with the current load characteristic; obtaining a real-time voltage / current regulation instruction set with a feed-forward compensation coefficient according to the dynamic mode parameter packet and the error tolerance data; analyzing the real-time voltage / current regulation instruction set, and combining the historical working condition similarity data to obtain an optimized PID coefficient matrix and a feed-forward gain vector; and executing the optimized PID coefficient matrix and the feed-forward gain vector, and generating and outputting a four-dimensional space trajectory signal which is completely matched with the target slope trajectory as a direct driving reference of a power supply power device. According to the embodiment of the invention, high-precision and low-jitter real-time slope tracking control can be realized, and the dynamic response speed and anti-interference robustness of a power supply system are improved.

The invention discloses an object recognition method and device, equipment and a storage medium, and relates to the technical field of machine learning, and the method comprises the steps: carrying out the feature similarity analysis of a target object feature of a to-be-recognized object and a class index feature group corresponding to a target object class, and obtaining target feature similarity data; the category index feature group is a clustering center feature corresponding to the positive sample object feature; the positive sample object is a sample object which is screened based on a feature density clustering result of a plurality of original sample objects and belongs to a target object category; based on the target feature similarity data and a preset feature similarity threshold, determining object category indication information of the to-be-recognized object; the preset feature similarity threshold value is obtained by performing threshold value adaptive adjustment based on a preset identification index threshold value and the sample feature similarity data; the sample feature similar data represents the feature similarity condition between the positive sample object feature of the positive sample object and the category index feature group. By using the scheme of the invention, accurate identification of a new category can be rapidly realized.

This disclosure discloses a method for monitoring the state of a chamber using a chamber monitoring device. The method may include: measuring the state of the chamber to generate multiple state data; calculating the similarity between each state data and reference state data to obtain multiple state similarity data; obtaining multiple state similarity patterns by matching processing segment data of each processed segment with the multiple state similarity data; and using two or more state similarity patterns to determine whether the chamber is operating normally or to predict the time point of abnormal chamber operation.

This invention discloses a method and system for track fusion of ships. The method includes: first, selecting radar data and ESM data of all targets before a certain time point; performing spatial and temporal registration of the radar data and ESM data of all targets; then, calculating the similarity between data pairs based on a nearest neighbor fast discrimination method, including: distance similarity between radar data pairs, carrier frequency similarity between ESM data pairs, and azimuth similarity between radar data and ESM data; then, determining whether the data pairs correspond to the same target based on the similarity between the data pairs; if the data pairs correspond to the same target, associating the data pairs, and finally performing weighted fusion of the associating radar data and ESM data. Compared with existing technologies, this method can solve the problems of sparse data in the early stage of monitoring and the existence of detection interference and monitoring errors during the process, thereby improving the efficiency of track fusion.

The application discloses a bank intelligent operation knowledge base implementation method and system based on a large language model and RAG. The method first extracts business terms from structured, semi-structured and unstructured multi-source heterogeneous multi-modal data of a bank, and then constructs a business dictionary and an ontology tree containing term similarity data. Then, the release agency weight, time effectiveness coefficient and penalty correlation degree of the regulatory provisions are quantitatively calculated to obtain an authoritative value and sort. Then, the pre-trained large language model is fine-tuned as a vector representation model using a double-constraint loss function, and the business data is vectorized and stored in a vector database. Finally, for user queries, the vector database is retrieved and combined with the regulatory provision sorting result for rearrangement, and a business response is generated after compliance verification. The method can improve the unstructured knowledge processing capability and retrieval accuracy of the bank, and realize efficient intelligent operation.

The invention relates to an exposure sequence adaptive optimization method and device for multi-view automatic measurement, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring an original exposure time sequence of a plurality of viewpoints; the original exposure time sequence comprises a plurality of exposure segments, and each exposure segment corresponds to one viewpoint; calculating similarity data among the exposure segments, and grouping the exposure segments according to the similarity data to obtain a plurality of initial exposure groups; for each initial exposure group, re-grouping the exposure segments at the boundary position to obtain a plurality of boundary grouping results, and calculating grouping cost data for each boundary grouping result; and selecting a target grouping result from the boundary grouping results according to the grouping cost data, optimizing the target grouping result according to the grouping cost data to obtain a plurality of target exposure groups, and calculating the exposure time of each target exposure group. The method can shorten the measurement period.

The application discloses an industrial CT image reconstruction method and device, electronic equipment and storage equipment, comprising obtaining projection image data and preprocessing to obtain target image data; based on the pre-trained U-Net variant lightweight network, the target image data is processed to obtain fine gradient direction prediction data and structure level similarity data, the U-Net variant lightweight network is trained based on the projection image data and the true value image data, and the loss function of the U-Net variant lightweight network comprises a structure similarity loss; based on the target image data, the fine gradient direction prediction data and the structure level similarity data, the target function is iteratively updated by using an alternating direction multiplier method to obtain reconstructed image data, and the target function is a high-order regularization function constructed based on the second-order total generalized variation and the structure level non-local mean of the iterative image data. The application can inhibit the streak artifact and photon noise of sparse images, so as to meet the requirements of fast scanning and sparse image reconstruction quality.

The present application discloses a method and apparatus for training a semantic vector extraction model and representing semantic vectors. The method for training the semantic vector extraction model includes: obtaining a conversation corpus from the same scenario as a downstream task corresponding to the semantic vector as a first pre-training text; performing secondary training on the trained BERT language network based on the first pre-training text to generate a BERT language network that focuses on nouns and verbs in the text; obtaining text classification data and generating a text similarity dataset based on the text classification data; and fine-tuning a multi-task classification network based on the text similarity dataset to generate a semantic vector extraction model.

The application provides a sea area digital twin method and system based on buoy and station data fusion, belongs to the related technical field of sea area digital twin, through constructing a multi-source heterogeneous marine data acquisition system, establishing a data quality evaluation mechanism and a similarity data compression mechanism, realizing intelligent fusion processing of buoy data and station data, solving the problem of low data fusion precision in traditional technology; through the self-organizing growth network structure of the marine dynamic response enhancement model, the neuron connection can be automatically adjusted according to the spatio-temporal variation degree of the training data, forming an optimal topological structure, effectively solving the defect that the traditional fixed parameter model cannot adaptively process complex marine environment; through the establishment of a dynamic parameter adjustment mechanism and a multi-algorithm adaptive switching strategy, the most suitable prediction algorithm is automatically selected according to the correlation characteristics of the marine data sequence, realizing high-precision digital reproduction of complex marine environment.

The application relates to the technical field of big data processing and network security, and discloses a ubiquitous network data deduplication system supporting similarity data ownership verification, which comprises an encryption module, a label generation module, a deduplication decision module, a challenge generation module, a proof calculation module, a verification module and an integrity evidence module; the system derives a high-entropy session key through two-dimensional chaotic mapping to encrypt plaintext data blocks, generates a similarity label with a threshold, and realizes efficient deduplication based on three-stage Hamming distance determination; the dynamic challenge-response mechanism is used to provide an auditable ownership proof for each download request, and the tree and block chain evidence technology are combined to realize multi-granularity checking of data integrity and rapid positioning of damaged blocks, so that the storage efficiency, security and usability are considered.

The invention provides a talent evaluation management method and system based on artificial intelligence, and relates to the technical field of artificial intelligence, and the method comprises the steps: 1, obtaining the resume data of a to-be-evaluated talent, carrying out the text recognition processing of the resume data, and generating a structured text data set; 2, extracting a keyword feature vector from the structured text data set, calculating a similarity value between the keyword feature vector and a preset talent evaluation keyword library, and generating initial similarity data; and step 3, constructing a three-dimensional point cloud data set based on the initial similarity data, wherein each data point comprises weight coordinate information of the keyword. According to the method, through structured processing, multi-dimensional feature analysis and dynamic correction, the combination of talent evaluation from objective standardization to three-dimensional dynamics is realized, and the comprehensiveness and accuracy of evaluation are improved.

This invention discloses a deep learning-based NLOS recognition method applicable to complex environments in multiple scenarios. First, the collected UWB channel impulse response data from multiple scenarios is transformed to obtain discrete multipath impulse response data. Multipath delay power data and multipath similarity data are then calculated to construct a new sample dataset, followed by data preprocessing. Next, a dual-input feature fusion deep learning NLOS recognition model based on multipath information is constructed to complete NLOS classification in UWB positioning systems. This deep learning-based NLOS recognition method, applicable to complex environments in multiple scenarios, can solve the problems of low recognition capability, poor generalization accuracy, and poor recognition effect in complex scenarios with limited data, thereby achieving high NLOS recognition accuracy and effectively improving UWB positioning accuracy in complex environments.

This invention relates to a method and system for detecting internal short circuits in a battery. The method includes acquiring constant current charging data of the battery under different charging cycles; the charging conditions are the same for each charging cycle; performing time alignment processing on the constant current charging data corresponding to each charging cycle to obtain a first feature curve; performing interval sampling on the first feature curve to obtain a second feature curve; calculating battery similarity data based on the second feature curve; and determining that the battery is in an internal short circuit state when the similarity data is greater than a judgment threshold. By collecting constant current charging data under different charging cycles, performing time alignment processing and interval sampling to obtain curves, and calculating battery similarity data under different cycles based on these curves, the battery is determined to be in an internal short circuit state when the similarity data is greater than a judgment threshold. This method is easy to execute online in real time in resource-constrained embedded battery management systems, does not rely on a specific battery model, and has strong stability in detecting internal short circuits in batteries.

The present application provides an evaluation method and a data generation method. The present application is to appropriately evaluate the accuracy of recognition results obtained by a recognition model. The evaluation method of work recognition accuracy includes: a result acquisition step of acquiring a recognition result obtained by a recognition model configured to recognize a work category of a work related to a work of a work person by inputting data related to the work of the work person; a correct answer acquisition step of acquiring a correct answer category corresponding to the recognition result and indicating an actual work category; and an evaluation step of evaluating the accuracy of the recognition result using similarity data, the similarity data being data in which the similarity of a plurality of work categories to each other is defined in advance and including the similarity between the work category of the recognition result and the correct answer category.

The application provides a financial message pushing method and device, relates to the technical field of artificial intelligence, and can be applied to the technical field of finance or other technical fields. The financial message pushing method comprises the following steps: determining user reading content feature similarity data according to a pre-constructed financial message graph and user reading content data; constructing a financial relationship network according to the user reading content feature similarity data and user financial social strength data; determining a group to which a target user belongs according to the financial relationship network, and pushing financial messages of the group to the target user. The application can improve the accuracy of financial message pushing and achieve the effect of thousands of people with thousands of faces.

The invention discloses a multi-modal image generation system and method based on AI image analysis, and relates to the technical field of image processing, definition evaluation is performed based on pixel gray value data of each region of a to-be-processed image, and each key region of the to-be-processed image is determined according to a definition evaluation result; obtaining an image health response index based on the density data of each region of the to-be-processed image, the image recognition result accuracy data after each region of the to-be-processed image is missing, and the similarity data of each region of the to-be-processed image; and obtaining a processing demand response index based on the definition evaluation result and the image health response index of each key area of the to-be-processed image, performing corresponding operation on each key area of the to-be-processed image according to the processing demand response index, and performing corresponding definition processing operation on each key area of the to-be-processed image. The accuracy of information transmission in the network is ensured, a large number of computing resources are saved, the data storage space is reduced, and the response speed and stability of the whole system are improved.

The application discloses a feedback data generation method, a question and answer system, a question and answer device and a refrigerator. The method comprises the following steps: receiving question data, performing fuzzy search and calculation in a feedback basic database according to the question data, obtaining a standard question sequence and at least two groups of target question sequences; performing bidirectional recurrent neural network traversal prediction and window sliding convolution on the standard question sequence and the target question sequence respectively, and correspondingly obtaining a standard question feature matrix and a target question feature matrix; performing projection fusion on the standard question feature matrix and the target question feature matrix, and assigning an attention vector to the standard question feature matrix and the target question feature matrix according to the obtained question judgment matrix, and comparing and calculating to obtain question similarity data; sorting the target question sequences according to the question similarity data, searching and outputting the final answer corpus of the target question sequence corresponding to the maximum question similarity data. The application can improve the accuracy and speed of feedback and simplify the architecture.

The invention relates to a decoupling double-speed self-adaptive test method and device, equipment, a medium and a product. The method comprises the following steps: acquiring a tested image sample and a template cue word; wherein the template cue word is set based on the category of the trained task; obtaining a pre-trained target decoupling double-speed adaptive model, and performing feature extraction on the image sample and the template cue word according to the target decoupling double-speed adaptive model to obtain image features and text features; wherein the pre-trained target decoupling double-speed self-adaptive model is obtained through a decoupling double-speed self-adaptive fine tuning step; performing similarity calculation on the image features and the text features to obtain similarity data; and testing the target decoupling double-speed adaptive model according to the similarity data to obtain a test result. By adopting the method, the plasticity and the stability of the model can be effectively balanced under the condition of data imbalance, and the test performance of the head class and the tail class is considered at the same time.

The present application relates to hyperspectral image processing technical field, disclose a kind of fast medical hyperspectral image classification method, comprising: extracting several pixels as training sample from the medical hyperspectral image to be classified;Set initial neighborhood window scale;According to initial neighborhood window scale, randomly select joint training area from the medical hyperspectral image to be classified;The similarity result between each pixel in joint training area and training sample is obtained by joint classification algorithm calculation;By k times iteration cross-validation algorithm, the initial neighborhood window scale is optimized, and the best neighborhood window scale is obtained;Using the best neighborhood window scale and joint classification algorithm, all pixels of the medical hyperspectral image to be classified are classified.Cotangent mapping cosine similarity can effectively reduce the interference of heterogeneous pixels, and local similarity data gravity can adaptively balance the contribution of minority and majority, to ensure the full use of MHSI spatial spectral information.

The application provides a method and device for determining fault type of a numerical control machine tool, wherein the method comprises: obtaining a first vector and a plurality of second vectors in a database, wherein the database is a database for storing knowledge of the numerical control machine tool in a manner of a knowledge graph; the first vector is used for representing an entity of the numerical control machine tool that has a fault, and the second vector is used for representing a fault type of the numerical control machine tool; calculating corresponding fault type similarities according to the first vector and the plurality of second vectors to obtain similarity data; and determining the fault type of the numerical control machine tool according to the similarity data. Through the application, the problem that the fault type is determined in a manner of manual intervention in the prior art, resulting in low fault maintenance efficiency, is solved.

The invention discloses a message response method and device, electronic equipment and a computer program product. Relates to the field of software testing, and comprises the steps that a request message sent by a tested system is received, M message elements are extracted from the request message, the request message is a message generated after the tested system analyzes a test case selected by a user, and M is a positive integer; the similarity between the M message elements and the N candidate elements is calculated, M * N pieces of similarity data are obtained, M target elements are determined from the candidate elements associated with the M * N pieces of similarity data, N is larger than M, and N is a positive integer; and forming a response message based on the M target elements, and sending the response message to the tested system. Through the message matching method and device, the technical problems of low matching accuracy and poor flexibility when the response message of the complex structured message is matched in the prior art are solved.

The present application relates to a kind of adaptive empathic dialogue methods based on quantitative emotional analysis, by using pre-training language model encoder, the dialogue sentence of user input is encoded and processed as context-related hidden state sequence, the sentiment contribution of each vocabulary in the dialogue sentence to dialogue history is quantitatively calculated, obtain global sentiment representation data, global sentiment representation data is respectively mapped to different polarity sentiment space and fine-grained emotional category space, obtain sentiment polarity probability data and emotional probability data;Global sentiment representation data, sentiment polarity probability data and emotional probability data are used to quantitatively calculate the emotional clues in current speech, the dialogue target is generated using large language model and the similarity data representing the matching of emotional clues and encoded dialogue target is obtained, the response strategy is optimized based on the similarity data, and the personalized empathic dialogue matching the current dialogue sentence of user is generated, accurately realizing the personalized adaptive empathic response to user emotion.

This invention provides a talent evaluation and management method and system based on artificial intelligence, relating to the field of artificial intelligence technology. The method includes: Step 1, acquiring resume data of the talent to be evaluated, and performing text recognition processing on the resume data to generate a structured text dataset; Step 2, extracting keyword feature vectors from the structured text dataset, calculating the similarity value between the keyword feature vectors and a preset talent evaluation keyword library, and generating initial similarity data; Step 3, constructing a three-dimensional point cloud dataset based on the initial similarity data, wherein each data point contains the weight coordinate information of the keyword. This invention, through structured processing, multi-dimensional feature analysis, and dynamic correction, achieves a combination of objective standardization and three-dimensional dynamics in talent evaluation, improving the comprehensiveness and accuracy of the evaluation.

The present application relates to the technical field of corn harvesting, and discloses a corn automatic harvesting system and method. The method counts a plurality of monitoring subareas of a corn planting area, deploys an image acquisition device, acquires growth characteristic data such as corn cob maturity distribution, stalk strength and leaf coverage density of each subarea, acquires real-time warehouse capacity data and conveying equipment operation state data of a warehouse scheduling center, determines initial harvesting parameters including harvesting mechanical advancing speed and corn cob stripping strength, judges whether corn cob maturity distribution data triggers a harvesting parameter adjustment condition, acquires plant image data of all subareas and extracts features to obtain plant morphological characteristic data when the condition is triggered, calculates subarea harvesting priority characteristic values according to the plant morphological characteristic data, matches the subarea harvesting priority characteristic values with a historical database, adjusts the initial harvesting parameters to obtain dynamic harvesting parameters according to a matching result, and calculates a similarity data set and corrects the dynamic harvesting parameters based on the similarity data set if there is no matching value in the historical database.

This application provides an identity recognition method, identity recognition device, electronic device, and storage medium, applied to augmented reality devices, belonging to the field of image processing technology. It acquires a first hand image, performs hand keypoint detection on the first hand image to obtain first keypoint position features, extracts a first region of interest from the first hand image based on the first keypoint position features, extracts palmprint features from the first region of interest to obtain first palmprint features, extracts palmprint features from a preset template to obtain second palmprint features, calculates similarity data between the first and second palmprint features, and performs identity recognition based on the similarity data and a preset similarity threshold to log in to the augmented reality device. This reduces the cost of identity recognition while achieving secure and fast identity recognition, thereby improving login efficiency and security.