38 results about "Granular computing" patented technology

The invention discloses a meta universe automatic driving obstacle identification method and system based on granular ball calculation. The method comprises the steps of extracting image space features; according to the local density of each feature point and a conditional purity criterion, carrying out particle-ball division; performing hierarchical aggregation on the pellets according to the spatial proximity relation and the feature similarity; taking each granular ball in the granular ball set corresponding to each image as a node, and determining edges among the granular balls according to spatial positions among the granular balls so as to construct a graph structure corresponding to each image; and inputting the node feature matrix and the adjacent matrix of the graph structure corresponding to the image into the trained graph neural network model to obtain a particle-ball-level obstacle recognition result, and mapping the recognition result back to the original image or the meta-cosmic space to realize pixel-level marking and classification of the target area.

The invention discloses a computing resource evaluation method for a meteorological large model. The method comprises the following steps: step (1), computing and evaluating parameters of the meteorological large model; step (2), calculating and evaluating the meteorological large model FLOPs; (3) calculating and evaluating the occupation of the video memory of the meteorological large model; and step (4), distributed communication evaluation of the meteorological large model. According to the invention, a multi-granularity computing resource joint evaluation framework is proposed for the first time, and the framework establishes a parameter quantity calculation model, a space-time perception FLOPs evaluation model, a video memory occupation model and a distributed communication analysis model through sub-modules and combines space-time heterogeneity characteristics of meteorological data to realize quantification of hardware resources by a meteorological large model.

The invention provides a power plant equipment sound abnormity identification and health prediction method based on granular computing and an LSTM network, and the method comprises the steps: employing an array pickup and a high-dynamic-range microphone array in a complex and high-noise background environment of a power plant, and combining an anti-interference filtering and beam forming algorithm, thereby achieving the sound abnormity identification and health prediction of the power plant equipment. According to the method, directional, multi-channel and non-contact sound acquisition is carried out on key parts of equipment, acquired sound signals are preprocessed, converted into time domain, frequency domain and time-frequency domain representations and coded into multi-dimensional numerical vectors, and compared with a rule-based expert system, the method has higher generalization ability and real-time response ability; a large language model is introduced, so that the output is closer to an engineering language and is suitable for operation and maintenance personnel to understand and execute; a self-defined knowledge base or safety semantic filtering is supported, and closed-loop deployment in an industrial field is facilitated; the system can be in butt joint with an intelligent inspection system, and full-link linkage of voice recognition, health assessment and suggestion generation is achieved.

The invention belongs to the field of federated learning technology application, and particularly relates to a personalized federated learning method based on a multi-granularity calculation unit, which comprises the following steps: taking a global client as a calculation unit at a coarse granularity level, calculating the membership degree of each cluster by each client, and then voting to select the cluster to which the client belongs, initially aggregating the clients with similar data distribution conditions together; a client cluster is used as a computing unit on a medium granularity level, and a cluster model and a client model are dynamically integrated; a single client is used as a computing unit on a fine-grained level, and implementation of personalized federated learning is accelerated by adjusting computing participation degrees of different layers of a neural network in a federated process, adopting a sparse weight activation training mode and executing sparse convolution in forward and backward propagation. According to the method provided by the invention, from the perspective of a multi-granularity calculation unit, under a data heterogeneous scene, the performance of a client model is effectively improved, and the calculation overhead of federal training is reduced.

The present invention discloses an abnormal learner detection method based on graph aggregation and restoration supported by granular computing, which can be applied to the field of graph aggregation technology. The method of the present invention includes the following steps: obtaining learner data of an online learning platform; performing data granulation on the learner data to obtain a learner relationship graph; inputting the learner relationship graph into a learner anomaly detection model to predict a learning anomaly detection result of the learner; wherein, the learner anomaly detection model is constructed through the following steps: constructing an aggregation layer by using a graph algorithm based on modularity; constructing a restoration layer by using a skip connection algorithm; constructing an output layer based on the aggregation layer and the restoration layer. The present invention can combine the influence of learner groups at different granularities on anomaly detection, and effectively improves the accuracy of anomaly detection.

The invention discloses a federated text open intention classification method based on granular ball calculation, which comprises a client training step, a server aggregation step and a client classification step which are completed in sequence, and is characterized in that the client training step uploads obtained pre-training language model parameters and a local granular ball knowledge base to a server; the server aggregation step constructs a global model and a global granular ball knowledge base, and issues the global model and the global granular ball knowledge base to each client; a client classification step: inputting a to-be-classified sample into the global model to obtain a global feature; and comparing the global features with a global granular ball knowledge base to realize intention classification. According to the method, the granular ball structured knowledge is constructed at the client side to realize efficient representation learning, the model and the knowledge are aggregated at the server side to support global reasoning and open recognition, and the method is suitable for a privacy protection scene of multi-client and multi-source heterogeneous text data; and accurate identification of known intentions and effective rejection of unknown intentions can still be realized.

The application provides a multi-modal collaborative evolution irony recognition method and device based on granular computing, and relates to the technical field of natural language processing. The method first extracts sample data from multi-modal, completes preprocessing and irony label labeling, forms a multi-modal irony data set aligned across modalities; then uses hierarchical granularity analysis to perform feature clustering in each modal feature space, constructs multi-modal multi-granularity knowledge representation; then filters a number of granularities with the highest dependency from each modality based on a rough set dependency function, generates a multi-modal optimal granularity matrix; then extracts the prototype mode vector of the known irony label and the test mode vector of the to-be-tested sample from the matrix, constructs an irony order parameter that characterizes the similarity between the two; finally, a collaborative neural network model is constructed based on the principle of synergetics, so that each modal order parameter evolves collaboratively under the mechanisms of self-excitation, self-inhibition and lateral inhibition, and the irony labeling mode corresponding to the highest order parameter is output after weight fusion, realizing irony recognition.

The application provides an information recommendation method, system, device and storage medium, comprising: by associating a special graph structure for different expert networks, the physical and logical decoupling of modal information is realized, and the embedding feature information of the target user and the candidate item in different semantic dimensions can be more accurately described. By calculating the data routing distribution with the user-item interaction instance as the granularity, and combining the progressive routing strategy with the prior routing distribution to obtain a hybrid routing distribution, accurate expert scheduling can be realized in different context environments, that is, it can automatically identify which modal features play a key role in user decision-making in a specific scenario, thereby accurately capturing the fine-grained preferences of users. Through the gating selection mechanism, the feature contribution of key experts is retained, effectively filtering the noise interference generated by irrelevant modalities, and further improving the accuracy of the prediction score and the reliability of the recommendation result.

The invention discloses a smart tourism recommendation method based on granular computing and a type-2 fuzzy set. The method comprises the following steps: acquiring multi-source heterogeneous tourism data; performing uncertainty modeling on the multi-source heterogeneous tourism data by using a Bayesian neural network, outputting posterior distribution, and mapping the posterior distribution into membership function parameters of a generalized type-2 fuzzy set to obtain multi-modal type-2 fuzzy information particles; based on a particle calculation framework, with coverage rate-specificity collaborative maximization as a target, performing optimal particle size distribution on the multi-mode type-2 fuzzy information particles to generate optimal particle size information particles; inputting the optimal granularity information particles into a two-channel deductive learning framework, realizing knowledge-data dynamic fusion by the two channels through a shared attention layer, and outputting a fused tourist preference feature vector; and based on the preference feature vectors, constructing a multi-granularity graph neural collaborative filtering model, respectively generating a personalized recommendation list, a group recommendation list and a socialized recommendation list, and outputting interpretable rules.

The invention discloses a resource semantic perception adaptive granularity network-on-chip atomic operation system and method, belongs to the technical field of network-on-chip, and aims to solve the technical problem of how to fundamentally reduce lock overhead, eliminate protection redundancy and improve resource utilization efficiency and chip concurrency performance through network-on-chip atomic operation. According to the technical scheme, the system comprises a resource semantic library, a plurality of on-chip network nodes, and a dynamic granularity generator, a semantic driving actuator, a source router or a target router which are deployed on the on-chip network nodes; the resource semantic database is used for storing semantic feature parameters of all competitive resources on a chip and providing a basis for granularity calculation and atomic operation strategy selection; and the on-chip network node is used for identifying an operation needing atomic protection according to upper-layer service logic, actively triggering an atomic operation request, packaging key information of the atomic operation into a request packet according to an NoC communication protocol, and sending the generated request packet to the source router through a network interface.

The application provides a virtual network function deployment method and a deployment device. The deployment method allocates resources for VDU required by a VNF to be deployed, takes the total VDU required as a whole group, calculates and schedules resources required by the total VDU in a calculation period in a group granularity, and until the whole group of VDU is scheduled to obtain corresponding resources or the whole group of VDU fails to be scheduled. The deployment position of the VDU of the VNF to be deployed is calculated in a group granularity, and the allocation mode not conducive to the overall deployment is recalculated and adjusted, so that the success rate of VNF deployment is improved.

The present invention belongs to the technical field of natural language processing, and particularly relates to a multi-granularity entity recognition method for pathological text naming. The method includes: obtaining pathological text information, and segmenting the pathological text at the character granularity and the word granularity; performing random mask masking and vector initialization on the segmented text, and using two parameter-sharing Bert models to encode the text after random mask masking and vector initialization; presetting a central replacement word and a central replacement character for each entity of each category; using KL loss and CE loss to construct loss functions for the character granularity and the word granularity, optimizing the CE loss for calculating the loss of the replaced character granularity, and optimizing the KE loss for calculating the loss of the replaced word granularity to obtain the entity recognition result. The present invention constructs templates through the character granularity and the word granularity for prediction, can accurately identify and extract the entities of the pathological text, and has a good entity recognition effect.

The invention discloses a health degree assessment method and device for site assets, electronic equipment and a storage medium, and relates to the technical field of software and platforms or other related technical fields, and the method comprises the steps: collecting multi-source data of site assets in a target site, and extracting a feature index value; calculating the health degree score of the site assets under the asset type according to the asset granularity calculation rule and the characteristic index value, and calculating the health degree score of the target site under the asset type according to the site granularity calculation rule and the characteristic index value; and constructing a health degree view of each asset type, inputting the health degree score of the site assets under the asset type and the health degree score of the target site into an evaluation model corresponding to the asset type, and outputting the transformation priority of the site assets in the site through the evaluation model. According to the method and the device, the technical problems of relatively low evaluation efficiency and relatively low evaluation accuracy of a mode of evaluating the health state of the site assets through manual inspection in related technologies are solved.

This application provides a virtual network function deployment method and deployment apparatus. According to the deployment method provided in this application, when resources are allocated to VDUs required by a to-be-deployed VNF, all of the required VDUs are used as an entire group, to compute, in one computation periodicity at a granularity of a group, resources required for scheduling all the VDUs, until the corresponding resources are obtained for scheduling the entire group of VDUs or scheduling of the entire group of VDUs fails. When deployment locations of the VDUs for the to-be-deployed VNF are computed at the granularity of a group, an allocation manner that is not conducive to overall deployment is recomputed and adjusted, so that a success rate of VNF deployment is improved.

An on-line analysis method for the dumping granularity of a mining truck comprises the steps that one or more cameras are installed above or obliquely above the dumping position of the mining truck, data transmission connection is established between the cameras and a central control room work station, and before on-line analysis is started, the cameras are connected with the central control room work station; the method comprises the following steps: firstly, obtaining a proportional relation among a shuffled ore area, a pixel size and a real object size through an offline shuffled image, and training an ore recognition model; during online analysis, cutting a real-time shuffling image according to a shuffling ore area, inputting the real-time shuffling image into an ore recognition model, outputting pixel shape detection information of each ore, and then performing granularity calculation on the image when the mining truck is in a normal shuffling state to obtain a real particle size length of each ore; then, the number of pixels in each piece of ore is fitted, and the proportion of ores of different types and fractions is counted according to the real particle size length of each piece of ore. According to the method, the ore granularity in the dumping ore card can be accurately analyzed online in real time under the complex background.

The invention discloses a dialogue generation method and system based on user portrait and mixed reward reinforcement learning, and the method comprises the steps: constructing a real user portrait based on real dialogue data; taking the distribution characteristics of the real user portraits as constraints, and generating corresponding virtual user portraits; taking the virtual user portrait as input, performing multi-round dialogue simulation through the dialogue interaction model after condition supervision fine tuning, and generating a virtual dialogue sample; based on the virtual dialogue sample, multi-dimensional rewards are calculated according to granularity and fused, and a mixed reward is obtained; and carrying out iterative updating on the dialogue interaction model by adopting a strategy optimization algorithm based on the mixed rewards. According to the method, various virtual user portraits are generated through potential spatial structure sampling, user coverage is expanded, and generalization ability is improved; and in combination with information entropy gain and other multi-dimensional mixed reward mechanisms, natural progressive expression of intentions and role stability are realized, and the authenticity, coherence and robustness of multiple rounds of conversations are remarkably improved.

The invention discloses an electricity utilization adjustment method and system based on matching of a network side green electricity proportion and electricity consumption, a storage medium and computer equipment. The method comprises the following steps: S1, collecting network side and load side data of a power grid; s2, calculating the grid-side green electricity proportion and the load-side electricity consumption of the power grid according to the unified time granularity; s3, carrying out per-unit processing on the grid-side green electricity proportion and the load-side electricity consumption of the power grid; s4, calculating the matching degree of the per-unit grid-side green electricity proportion and the load-side electricity consumption; s5, judging whether the matching degree reaches the standard or not, if the matching degree does not reach the standard, adjusting the electricity consumption of the load side according to the difference value between the per-unit grid side green electricity proportion and the electricity consumption of the load side in the adjacent time periods, performing per-unit processing on the adjusted electricity consumption of the load side again, and then returning to the step S3; if yes, adjustment is ended; the green electricity utilization rate can be improved by adjusting the electricity consumption of the load side, and carbon reduction is achieved.

The invention discloses a block chain and interpretable large model driven climate investment and financing method, the method combines a block chain and an interpretable large model, on one hand, the upper chain management of project data and the execution of an intelligent contract are realized, the whole financing process is ensured to be open, transparent and non-tampered, and through the automatic process design of the intelligent contract, the financing efficiency is improved. Manual intervention is reduced, and efficient and credible fund management is realized; and on the other hand, pre-training and fine tuning are performed based on a large model technology, so that the large model is good at processing climate type projects and ensures that an output format meets requirements, the accuracy and reliability of large model risk assessment are improved, the loss caused by inaccurate risk assessment is reduced, and meanwhile, the risk assessment efficiency is improved. According to the method, the explainable risk index based on the granular computing theory and the expert experience is constructed, the explainable risk index and the risk index output by the large model form complementation to form a double-risk mechanism, the large model output is explained and adjusted, the illusion phenomenon of the large model is reduced, and the predictive capacity and the explainability of the large model are effectively balanced.

The invention discloses a bit stream, a bit stream signature and an authentication method, and relates to the technical field of multimedia. The bit stream signature method comprises the steps that the computing device obtains authentication data and then outputs a bit stream, and the bit stream comprises the authentication data. Wherein the authentication data comprises signature data, the signature data is obtained by performing signature according to abstract data of each layer unit in a group of layer units of the bit stream, and one layer unit in the group of layer units comprises a network abstraction layer (NAL) unit with the same layer identifier in the bit stream. The abstract data is calculated by taking the layer unit as granularity, so that only the corresponding layer unit fails under the conditions of frame loss, packet loss or layer unit authentication failure and the like in the transmission process. If only the layer unit which fails to be authenticated loses efficacy, that is, the lost data is less, the utilization rate of the data in the bit stream can be improved, and the transmission efficiency is improved.

The present invention discloses a method, device, and medium for constructing prompt words based on domain knowledge for granular computing. The method includes the following steps: obtaining domain knowledge and constructing prompt words according to the domain knowledge; inputting the prompt words into a large language model to obtain an attribute pair probability matrix; judging the correlation between particles to be calculated according to the attribute pair probability matrix; when the correlation between the particles to be calculated meets the requirements, performing a granular computing operation on the particles to be calculated. In the embodiments of the present invention, constraint conditions are constructed during the granular computing process through domain knowledge, which can reduce the discovery of invalid object patterns during the granular computing process and obtain object patterns with higher correlation as inference results with higher quality. The object patterns obtained in the embodiments of the present invention can better prompt the attribute information that industry personnel need to focus on, provide a better guiding role for problem-solving, and are widely applied to the granular computing solution process in different industry fields.

The invention provides a spatio-temporal data encoding and decoding method and device based on granular computing, and relates to the technical field of computer big data processing. The method comprises the steps of performing compression processing and normalization processing on spatio-temporal data to obtain processed spatio-temporal data; coding the processed spatio-temporal data by using a spatio-temporal enhanced maximum entropy fuzzy coding algorithm to obtain a plurality of coding centers and a plurality of membership matrixes, determining a plurality of decoded spatio-temporal data according to the plurality of coding centers, the plurality of membership matrixes and the fuzzy coefficient, and decoding the decoded spatio-temporal data to obtain a plurality of decoded spatio-temporal data; determining a plurality of errors corresponding to the plurality of decoded spatio-temporal data and the processed spatio-temporal data; and selecting a minimum error from the plurality of errors, and taking the target membership matrix and the target coding center corresponding to the minimum error as a target coding result. Thus, the processed spatio-temporal data can be coded through the spatio-temporal enhanced maximum entropy fuzzy coding algorithm, the time required for processing massive spatio-temporal data is greatly shortened, and the spatio-temporal data processing efficiency can be improved.

The present application relates to the technical field of distributed storage. Disclosed are a metadata addressing method and apparatus, and a medium and a product. A target number of metadata service paths are obtained by filtering metadata feature parameters within a preset time, such that local management is performed on metadata in a large-scale distributed storage file system, and path information of the metadata is recorded and counted to improve the locality management of the metadata; moreover, a first path result is obtained by means of granularity calculation, and granularity selection is performed on the metadata paths to improve the temporal locality of the metadata; in addition, a second path result is obtained by evaluating and calculating the first path result on the basis of a path optimization strategy, such that locality processing is performed on the second path result on the basis of a locality function. Thus, the locality data structure of metadata is changed, thereby ensuring the spatial locality of metadata storage paths in a distributed file system, optimizing the paths for metadata operations in the distributed storage file system and improving the I / O performance of a storage system.

The application discloses a self-adaptive neighborhood granular clustering method suitable for mixed attribute data and belongs to the technical field of machine learning. In view of the problems that the traditional K-means clustering algorithm can only process numerical attribute data and the clustering result is unstable in the prior art, the application provides a self-adaptive neighborhood granular clustering method suitable for mixed attribute data, which comprises the following steps: S1, obtaining and inputting to-be-clustered data; S2, calculating neighborhood granular vectors of the to-be-clustered data by the self-adaptive neighborhood granular clustering method of mixed attribute data; S3, selecting initial clustering centers of the to-be-clustered data based on a dissimilarity measurement method of data; S4, updating the clustering centers by using a neighborhood granular K-means clustering algorithm; and S5, outputting a clustering result. Thus, the K-means clustering algorithm and granular computing are combined, mixed attribute data granulation is realized, and the self-adaptive neighborhood granular clustering method has the excellent characteristics of high applicability and high clustering performance.

The application discloses a kind of federal text open intention classification methods based on granular computing, including sequentially completed client training step, server aggregation step, client classification step, and the obtained pre-training language model parameter and local granular ball knowledge base are uploaded to server in client training step;Global model and global granular ball knowledge base are constructed in server aggregation step, and are issued to each client;Global feature is obtained by inputting the sample to be classified into global model in client classification step;And global feature is compared with global granular ball knowledge base, and intention classification is realized.The application constructs granular ball structured knowledge in client to realize efficient representation learning, and aggregation model and knowledge are supported in server side to support global reasoning and open identification suitable for privacy protection scene of multi-client, multi-source heterogeneous text data, under the premise that user data cannot be centrally summarized, still can realize accurate identification to known intention and effective rejection to unknown intention.

The invention discloses an electric power data transmission method and device based on Hash slicing, terminal equipment and a storage medium, and belongs to the technical field of data transmission, the method comprises the following steps: obtaining original electric power test data of electric power equipment, and carrying out division and Hash operation according to a preset time window to obtain a Hash segment with a Hash number; calculating a first segmentation position of the Hash segment of the non-sudden deformation waveform according to a preset slice granularity, calculating a second segmentation position of the Hash segment of the sudden change waveform according to a preset sudden change detection function and a preset sudden change threshold, and taking a key frame in the Hash segment of the video stream as a third segmentation position, after segmentation, obtaining a plurality of data slices and a slice number of each data slice; and finally, transmitting the data slices according to the preset priority weight of each data slice. By implementing the method and the device, the problem of low data processing efficiency during data transmission caused by a fixed and single data segmentation mode in the prior art can be solved.

The invention belongs to the field of automatic driving environment perception, and relates to a clustering roadside perception message comprehensive filtering method applying granularity calculation, which comprises the following steps that: vehicle end equipment acquires data information of a roadside unit, and preprocesses the data information; respectively extracting n messages from the processed data information, and forming a sample point set by the n messages; dividing samples in the sample point set; screening out a candidate set according to the divided samples; constructing a common matrix according to the candidate set; fusing the data in the candidate set by adopting a hierarchical clustering algorithm based on condensation on the basis of the common matrix to obtain Pf; calculating a contour coefficient of the Pf, and screening the data according to the contour coefficient; combining the screened data to obtain a minimum non-repeated set; according to the invention, the problem of information redundancy caused by imperfect deployment strategies of RSU roadside equipment and vehicle end acquisition equipment, complicated equipment performance and traffic environment, and shielding of tall buildings and trucks is reduced.

Disclosed are a bitstream, a bitstream signing method, and a bitstream verification method, relating to the technical field of multimedia. The bitstream signing method comprises: a computing device acquires verification data, and then outputs a bitstream, the bitstream comprising the verification data. The verification data comprises signature data, which is obtained by signing on the basis of digest data of each layer unit in a group of layer units of the bitstream, and one layer unit in the group of layer units comprises network abstraction layer (NAL) units having a same hierarchical identifier in the bitstream. By means of computing the digest data by using layer units as the granularity, only a corresponding layer unit is invalidated when frame loss, packet loss, or layer unit verification failure occurs during transmission. When only the layer unit for which verification failure occurs is invalidated, that is, less data is lost, the utilization rate of data in the bitstream can be increased, and the transmission efficiency is improved.

The invention provides a long-term PM2.5 concentration prediction method based on granular computing and parallel LSTM, and belongs to the technical field of environmental data prediction and artificial intelligence application. The method comprises the following steps: preprocessing an original PM2.5 concentration time sequence; searching an optimal segmentation scheme by using a particle swarm optimization algorithm, and segmenting the time sequence to obtain a particle sequence; constructing three mutually independent LSTM neural networks, and performing training by taking a mean value sequence, a standard deviation sequence and a sample point number sequence in the particle sequence as input; the three trained LSTM neural networks are used to predict future features, and M prediction triples are obtained; and constructing Gaussian distribution for each prediction triad, generating prediction data points corresponding to each prediction triad, and splicing the prediction data points according to a time sequence to form a complete future PM2.5 concentration prediction sequence. According to the method, the problems of error accumulation and time sequence correlation weakening can be relieved, and the prediction accuracy is improved.

The invention discloses an action recognition model lightweight method based on hierarchical dynamic fusion network pruning. The method comprises the following steps: uniformly sampling an original video to obtain an RGB frame sequence, and constructing an initial action recognition model; designing a multi-granularity computational graph coding module, weighting and randomly selecting pruning granularity according to the redundancy of the current granularity network parameters, and dynamically describing an interlayer parameter relationship by utilizing the frequency domain similarity of each granularity parameter output feature graph, so as to generate a corresponding hierarchical graph structure; thirdly, searching a key path in the graph by utilizing a self-adaptive pruning fusion module to dynamically prune redundant parameter nodes; and finally, performing fine tuning optimization on the pruned action recognition model to obtain a lightweight action recognition network. According to the method, the capability of describing time sequence dependence by network parameters can be evaluated, the pruning position, the pruning granularity and the pruning proportion can be adaptively adjusted, the reasoning speed of the action recognition model is increased, the quantity of model parameters is reduced, and lightweight deployment of the action recognition model is facilitated.