25 results about "Clustering search" patented technology

The invention relates to the technical field of glue pudding quality detection, and discloses a glue pudding quality intelligent detection system and method based on multi-sensor fusion. The system comprises a data acquisition module, a stability analysis module, a detection parameter search module, a quality detection execution module and a quality judgment module. The data acquisition module acquires a surface image, weight and texture elasticity data set of the glue pudding sample through a multi-sensor array; the stability analysis module evaluates the acquired data to obtain image, weight and texture elastic stability factors; the detection parameter search module takes each stability factor as a joint index, and determines an optimal detection parameter group through density clustering search in a detection parameter space; the quality detection execution module detects the target glue pudding of the production line according to the optimal parameter group to obtain multi-source detection data; and the quality judgment module inputs the multi-source data into a classifier to obtain a quality grade judgment result. The system realizes comprehensive detection of the glue pudding quality, and improves the detection intelligence and automation level.

In an embodiment, operations include receiving a dataset including a set of graphs, applying a graph explainer model on the dataset to determine explain scores, and filtering graph elements of each graph based on the explain scores to determine a first set of graph elements. Further, another dataset of randomized graphs is generated. The graph explainer model is applied on the other dataset to determine noise scores, which are used determine denoised explain scores from the explain scores. Graph elements of each graph are filtered based on the denoised explain scores to determine a second set of graph elements. Thereafter, feature vectors are determined for the first set of graph elements and the second set of graph elements, and prediction driving graph elements are determined based on the feature vectors. Explainability information associated with the set of graphs is rendered, based on the prediction driving graph elements.

The invention relates to the technical field of intelligent power grids, and particularly provides a power grid intelligent decision-making method, system and device, a medium and a computer program product, and the method comprises the steps: dividing an intelligent agent cluster suitable for a target power grid; establishing an equivalent model of the agent cluster, and generating a decision model of an independent agent; in response to a power regulation and control instruction issued by the power grid coordination center, the target agent cluster searches a matched candidate agent cluster in the regional hierarchy; and the independent intelligent agent determines a matched target intelligent agent, and performs power adjustment according to the energy output information of the target intelligent agent in a preset historical duration. According to the technical scheme provided by the invention, the overall scheduling flexibility of the power grid system can be improved through the intelligent power grid system.

Methods and apparatuses for automatically generating and displaying clustered search results and contextual recommendations are described. While a user is making edits to an electronic document or within an application, a search system may identify a set of search terms based on information entered by the user and / or the locations of the information entered by the user within the electronic document or application and then generate and display search results based on the set of search terms. The search system may cluster search result documents that have a degree of similarity and display multiple clusters of search results to the user. The search system may allow users to specify personalized keyword links and may return a personalized keyword link as a search result for a given search query even if a document or resource referenced by the keyword link has not been indexed or crawled.

The invention relates to a KAN-based unmanned vehicle trajectory prediction method. The method comprises the steps of 1, screening unmanned vehicle threat distribution diagrams obtained according to unmanned aerial vehicle cluster search information to obtain a target probability distribution diagram; 2, extracting data set features through a high threat region weighted midpoint method or a clustering-based position extraction method to obtain a target position; and 3, selecting a proper activation function, and predicting a target trajectory by using the double-KAN prediction model. Compared with an existing trajectory prediction algorithm, the method has the advantages that feature extraction is carried out on the target threat distribution map, and the explicit expression is obtained through KAN training, so that the interpretability of target and trajectory search of the unmanned aerial vehicle cluster is further improved.

This invention discloses an energy-saving control method for pump sets based on discrete operating point clustering search optimization, belonging to the technical field of pump set scheduling control. It acquires pump set operating data based on sensors, predicts the target flow rate based on historical flow time-series data, and determines the target head. Under the premise of meeting water supply constraints, the frequency of any one or more pumps in the pump set is perturbed, and discrete operating points under different frequency combinations are collected. Clustering search optimization is performed based on the discrete operating points to obtain candidate clusters that meet the target requirements. Within the candidate clusters, a multi-index comprehensive evaluation model is constructed to score the points within the clusters, and the points within the clusters with higher scores are selected as representative points, obtaining a set of representative points. Based on the set of representative points, the representative point that best meets the target requirements and has the lowest adjustment cost is selected as the optimal operating point. This invention directly uses sampled points for optimal solution search, avoiding fitting errors and more closely reflecting actual operating conditions, thus possessing good practicality.

The application discloses a ruled surface and plane approximation method of NURBS surface, comprising: carrying out equal parameter line discretization and surface slicing pretreatment on a target NURBS surface; adopting a straight line cluster search technology based on principal component analysis method to obtain an initial straight line cluster approximating the NURBS surface; utilizing the initial straight line cluster to construct an initial ruled surface, and carrying out overall optimization on the initial ruled surface based on a least square method to obtain one or a group of ruled surfaces meeting the approximation error requirement. The application converts a complex NURBS surface into one or a group of ruled surfaces through three steps of pretreatment, straight line cluster search and overall optimization, and is suitable for efficient rough machining and tool path planning.

The invention discloses a three-dimensional track reconstruction method and device for muon imaging, and relates to the field of signal detection and ray imaging, and the method comprises the steps: obtaining an original waveform signal of muon imaging, and carrying out the preprocessing of the original waveform signal; performing channel waveform analysis on the preprocessed original waveform signal by adopting a waveform identification technology to obtain a channel waveform analysis result; in combination with a detector-electronics channel mapping table, decoding continuous channel information by utilizing hit continuity characteristic channel clustering; on the basis of the continuous channel information, utilizing a spatial point clustering algorithm to take spatially adjacent electronic signals in the continuous clusters as the same electronic cloud cluster; and based on the three-dimensional barycentric coordinates of the plurality of electronic clouds, completing three-dimensional track reconstruction by using a screening reconstruction algorithm. According to the method, the problem of fuzzy reconstruction of a small-angle incident muon example is solved through channel waveform analysis, hit cluster searching and three-dimensional track reconstruction.

The application discloses an optimal flyby sequence acquisition method based on a parallel architecture and belongs to the field of orbit design and optimization. The application realizes the method as follows: a parallel cluster search algorithm is constructed, and a Lambert problem is solved in parallel; a state of a spacecraft at a starting moment of a task time window is determined, a parallel cluster search is performed to obtain a sub-sequence of target access; then, the sub-sequence is saved to a database, a state and a moment at an end of each sub-sequence in the database are extracted, and the parallel cluster search is performed again from the state; the above process is repeatedly iterated to generate a sub-sequence database; the sub-sequences are spliced to generate a complete target access flyby sequence trajectory; and the total speed increment of the flyby sequence is further reduced by adjusting a transfer time between targets and optimizing a flyby speed. The application can significantly improve flyby sequence planning efficiency, effectively guarantee the optimality of sequence generation for a large-scale target problem, and acquire a high-precision and low-energy-consumption spacecraft flyby sequence.

The invention relates to the field of data processing, and relates to an e-commerce landing page intelligent generation method and system, and the method comprises the steps: obtaining a uniform resource locator of a target commodity, and collecting the multi-modal data of the target commodity based on the uniform resource locator; performing feature extraction on the multi-modal data to obtain structured style feature data including a visual feature vector and a semantic feature vector; a design space generated by the landing page is constructed, the design space is modeled into a directed graph structure, nodes represent page component modules, and edges represent connection relations between adjacent component modules; and performing path search in the directed graph structure by using a cluster search algorithm to obtain an optimal design scheme, and generating a target e-commerce landing page based on the optimal design scheme. By means of the method, the efficiency and quality of landing page generation can be effectively improved.

The application provides an AI image recognition front-end code generation method and system and a computer storage medium. The generation method includes three main steps of image recognition, layout analysis and code generation. The original webpage prototype and its configuration information are processed by a convolutional neural network, an optical character recognition technology and a target detection algorithm to extract interface elements and determine the hierarchical structure. Then, attribute information extraction and rendering tree generation are performed. Finally, data tokenization processing is performed to map into HTML, CSS or JavaScript code, and a greedy search or clustering search strategy is flexibly selected according to user requirements to generate the final front-end code. The system supports multiple front-end frameworks, ensures efficient and accurate code generation, and allows user interaction to enhance the individuality and adaptability of code generation. The application relates to the field of artificial intelligence technology.

A feature information comparison system includes a feature information extractor configured to extract feature information from each of a plurality of multimedia data, a quantization processor configured to quantize the extracted feature information and store it in a feature information database, a clustering processor configured to cluster a set of target multimedia data using the feature information, calculate centroid for each cluster, and store the centroids in a cluster database, a cluster searcher configured to compare the quantized feature information extracted from a search input with the cluster centroids in the cluster database, select similar clusters based on the comparison, and designate the multimedia data in the selected clusters as search candidates, and a search result provider configured to compare the feature information of the multimedia data in the selected clusters with the feature information of the search input and provide a final search result based on the comparison.

This invention discloses a clustering search method and system under time-series constraints, comprising: acquiring raw time-series data and constructing a time-series graph based on the raw time-series data; initializing the node attributes of each node in the time-series graph; traversing the nodes and determining whether a node is similar to its neighboring nodes; if similar, updating the similarity of the node and the similarity of its neighboring nodes; using the degree to indicate the number of neighboring nodes of a node; connecting neighboring nodes with nodes through edges; identifying nodes with a similarity not lower than a threshold for the number of core nodes as core nodes, and clustering based on the temporal structure similarity with the core nodes as the center to obtain structural clustering results. The above scheme can quickly locate and return complete clustering results related to the search target from the clustering structure when a search condition is received, thus improving search efficiency.

The application discloses a power transmission section searching method based on variable scale near neighbor propagation clustering, and steps of the method comprise the following steps: 1, establishing a power network model; 2, searching a power transmission section by variable scale near neighbor propagation clustering. The application considers the demand of the power network, divides the whole power network in different scales by performing multiple variable scale partition operations on a power topology diagram, searches more potential power transmission sections, is beneficial to a dispatching department to timely adjust an operation mode, thereby can avoid large-area power grid outage accidents, and guarantees safe and stable operation and optimal dispatching of the power grid.

The application discloses a maximum bipartite cluster search method and system based on local and global structure feature difference measurement, calculates the chi-square value of nodes according to input bipartite graph data; maps the chi-square value of two end nodes to adjacent edges; constructs a search space according to the large edge in front of the chi-square value; executes a maximum bipartite cluster search algorithm on the search space, and obtains a result. The application can effectively model based on local structure and global structure feature difference, and take the local structure with large difference as a heuristic search space, and then obtain a result through an optimized maximum bipartite cluster search algorithm. The method not only has high search efficiency, but also can improve 1-3 orders of magnitude compared with existing heuristic methods, and the result is better, and the method has strong adaptability and can meet the maximum bipartite cluster search demand in various scenes.

The invention relates to a scaling experiment method and system based on an unmanned aerial vehicle search algorithm, belongs to the field of virtual simulation, and solves the problems of high cost and high risk of full-size offshore experiment and insufficient credibility of pure digital simulation in the existing method. Comprising the following steps: scaling prototype parameters of an offshore full-size scene to a lake surface experiment scene according to a selected scaling coefficient; in the lake surface experiment scene, the unmanned aerial vehicle cluster executes a search task on a simulation target based on a predetermined collaborative search algorithm; task data of each round of search task executed by the unmanned aerial vehicle cluster in the lake surface experiment scene is collected, and lake surface performance data is obtained; and deducing the lake surface performance data to an offshore full-size scene to obtain a performance evaluation result of the collaborative search algorithm in the offshore full-size scene. Through a scaling experiment with low cost and high controllability, the performance of the unmanned aerial vehicle cluster search algorithm in a full-size sea area can be effectively and reliably predicted and evaluated, and the research and development cost and risk are remarkably reduced.

The technical scheme of the present application discloses a real-time closed-loop detection framework and an ultra-fast descriptor generator based on FPGA, characterized by comprising a descriptor generation module and a retrieval verification module; the descriptor generation module comprises a preprocessing submodule, a clustering submodule and a descriptor calculation submodule which are sequentially cascaded; the retrieval verification module comprises a retrieval submodule based on a bag-of-words method and a verification submodule based on global descriptor cross-validation. Compared with the prior art, the clustering search space of the present application is reduced by about 2000 times. The present application eliminates the blocking problem caused by data dependence through a non-blocking architecture, and reduces the descriptor calculation complexity through a cylindrical filter and fast direction calculation. During retrieval, the technical scheme disclosed by the present application only needs 5 local descriptors to maintain high accuracy, realizes a performance improvement of 128.8 times, a single-frame processing time of 16.5 milliseconds, meets the real-time requirement of 50 milliseconds, and the energy consumption is only 21.6 mJ / frame.

The application provides a UAV cluster distributed area search method and system based on role differentiation, initializes a UAV cooperative search task environment, updates a cluster search cognitive map by an information UAV, models task allocation of the UAV cluster cooperative search, pre-allocates a task based on CBAA, updates a cluster topology connectivity state of an information sub-cluster, allocates a task of a relay UAV based on connectivity maintenance, controls a UAV waypoint tracking based on collision time cooperative guidance, adjusts an information-relay sub-cluster ratio based on area coverage, and outputs a UAV cluster cooperative search result. The application makes the CBAA algorithm support connectivity maintenance distributed decision of the UAV cluster through improvement strategies such as unconnected individual removal, action selection taboo list and mobile base station switching. Meanwhile, the heterogeneous cooperation between the information-relay sub-cluster is utilized to release the action space of the UAV, enhance the persistence exploration ability of the cluster in the later period, and avoid local optimum of the search process.

An effective stock keeping unit (SKU) management system encodes catalog data into an embedding per catalog item. An embedding space is created by encoding catalog item data into an embedding per catalog item. The embedding is created by generating an index, where a number of rows represents a number of catalog items and a number of columns represents a number of fields associated with each catalog item. The index is then denormalized using customer groups and transformed by compressing the number of columns, to create the embedding space. In some configuration, a machine learning model is trained using catalog data. In the embedding space, item similarity is encoded by clustering catalog SKUs into groups in the embedding space, by placing similarly related items close to each other in the embedding space. Catalog items are then searched for in the embedding, with the closest clusters searched for a particular catalog item.

This invention belongs to the field of cluster analysis technology, specifically relating to a method for identifying malfunction patterns in relay protection based on cluster analysis. The method includes: collecting fault waveform data, extracting current surges, the proportion of zero-sequence components, and the second harmonic ratio, and constructing a feature set after normalization; calculating the local sparsity of samples based on the KNN algorithm, and constructing a density adjustment factor by combining the logarithmic difference of sparsity with an energy gating term; generating a dynamic radius using the density adjustment factor to replace the fixed search threshold in the clustering algorithm, and introducing shared nearest neighbor verification to remove pseudo-connected samples to construct a time-series cluster; based on the statistical characteristics of the time-series cluster, if the second harmonic ratio exceeds the limit, it is determined to be an inrush current and blocked; otherwise, a metallic or high-resistance grounding fault is determined based on the current surge and the duration of the zero-sequence component proportion. This invention solves the problems of difficulty in capturing sparse fault features and difficulty in identifying malfunctions by dynamically adjusting the cluster search range.

The application belongs to the technical field of computers, and particularly relates to a method and system for optimizing CPU overhead in the operation of a bundle search algorithm. The method for optimizing CPU overhead in the operation of a bundle search algorithm comprises the following steps: performing slice processing on the probability distribution set of each modeling unit; using an algorithm on a GPU to solve the top k candidate solutions with high probability; and splicing the slices of the top k candidate solution data set to output a complete probability distribution data set. After optimization of the bundle search algorithm, the CPU overhead is significantly reduced, and the occupation time of the GPU is also reduced. After optimization, the average time consumption of the entire model training process is improved by about 36%, and the optimization effect is obvious.

A preliminary tag search improves computer functioning by reducing cross-cluster searches. Today's cloud service providers maintain large, distributed datasets stored by multiple clusters having many nodes. Many cross-cluster searches generate no search results, but unsuccessful cross-cluster searches still consume much hardware resources, network bandwidth, and electrical energy. The preliminary tag search eliminates unfruitful cross-cluster searches. The preliminary tag search may identify only the cross-cluster searches that will generate positive search results. The preliminary tag search may identify wasteful cross-cluster searches that would generate no search results. Prior to conducting cross-cluster searches, a distributed database service may first perform the preliminary tag search that identifies successful / unsuccessful cross-cluster searches. The distributed database service may then decline or skip unsuccessful cross-cluster searches.

The invention relates to the technical field of intelligent earthquake rescue, in particular to an earthquake search and rescue robot combined search and rescue method and system based on swarm intelligence, and the method comprises the steps: constructing a distributed communication network and a role dynamic distribution mechanism of an earthquake search and rescue robot cluster; using an improved particle swarm optimization algorithm to realize cluster global path planning and regional collaborative coverage; life signal and dangerous area identification is completed through multi-robot heterogeneous sensing data fusion; carrying out search and rescue information real-time sharing and decision collaboration based on a group consensus mechanism; and a feedback regulation mechanism is introduced to dynamically optimize the cluster search and rescue strategy. Compared with the technical problems of low single-machine operation efficiency, limited coverage and serious information island of the earthquake search and rescue robot in the prior art, the cooperative closed loop of the earthquake search and rescue robot cluster is realized through swarm intelligence, and the search and rescue efficiency, the coverage integrity and the target recognition accuracy in the earthquake ruins environment are improved.

The application discloses a kind of global sequence perception-oriented product service combination fair recommendation method, belong to intelligent service computing and recommendation field.The method includes: introducing time bias to obtain TiTE encoder, according to the user set and product service set obtained by acquisition, the historical service call sequence of each user and historical service call timestamp sequence are passed through TiTE encoder, obtain hidden state sequence, further through self-recurrence and cluster search, generate multiple candidate service combinations;Each candidate service combination is filtered in turn according to two conditions of entity check and resource constraint, and the final score of each candidate service combination is calculated by combining interest score, throughput score and user hunger value;Establish joint programming model about final score, and the optimal candidate service combination of each user is obtained by using hierarchical greedy heuristic algorithm.The method of the application realizes the leap from single user local optimum to system overall efficiency optimum.