62 results about "Group model" patented technology

The invention relates to a multi-model-question-based group consensus large model illusion reduction method, which comprises the following steps of: screening a Top-N model from a candidate model set according to a multi-dimensional comprehensive scoring result, and executing full-combination bidirectional knowledge distillation on the Top-N model to obtain an initial model group, loading a plurality of domain knowledge bases for each model in the initial model group to carry out domain self-adaptive fine tuning, and constructing to obtain a group model set; giving a user question, triggering a plurality of fine-tuned field expert models in the group model set to perform parallel reasoning to generate an initial answer, performing iterative optimization by constructing a question set, generating a question instruction and updating the answer, and calculating the similarity of group answers by adopting a mixed kernel function in the iterative optimization process to obtain a group answer set; and when the similarity and the stability reach preset threshold values at the same time or reach the maximum number of iterations, stopping iteration and outputting a result. Compared with the prior art, the method has the advantages of high answering accuracy, high field adaptability and the like.

The invention discloses a social group simulation method and device based on multi-agent driving. The method comprises the following steps: performing user attribute sampling processing on a real e-commerce user data set, and constructing agent basic attributes based on user attributes in combination with a large language model to obtain agent basic attribute features; performing construction processing on the agent basic behaviors based on a preset memory mechanism in combination with the agent basic attribute features; according to the agent basic attribute features and the agent basic behavior features, performing construction processing on agent interaction behaviors in combination with a large language model; performing relation network construction processing based on a small-world network model on the plurality of agents; and performing simulation processing based on execution agent behaviors on the agent social group model to obtain a multi-user behavior simulation result. By constructing a multi-agent simulation system in an e-commerce scene and combining the multi-agent simulation system with a large language model, social group simulation in a complex e-commerce interaction scene is realized, and the accuracy of social group simulation in the e-commerce scene is improved.

The systems and methods disclosed herein receives artifacts generated using a first set of models within a multi-model superstructure. The multi-model superstructure includes a second set of models to test the first set of models. The multi-model superstructure dynamically routes the artifacts of the first set of models to one or more models of the second set of models by (i) determining a set of dimensions of the artifacts against which to evaluate the artifacts and (ii) identifying the models in the second set used to test the particular dimension. The second set of models then assesses each artifact against a set of assessment metrics. If an artifact fails to meet one or more assessment metrics, the second set of models generates actions to align the artifact with the set of assessment metrics.

A federated learning method is provided, applied to the field of artificial intelligence technologies. According to the method, when obtaining models of different network structures, an aggregation node groups models of a same network structure into a same group, and performs parameter aggregation on models in a same group, to obtain a plurality of aggregation models of different network structures. In addition, for each aggregation model, knowledge distillation training is performed on each aggregation model based on the plurality of originally obtained models, to implement experience transfer between the models of different network structures, so as to integrate knowledge and experience of models of various network structures, combine advantages of parameter aggregation and knowledge distillation in integrating model experience, implement aggregation of the models of different network structures, and ensure prediction precision of a model obtained through aggregation.

The invention discloses a cloud edge federal learning method and system and a storage medium, and belongs to the field of model training optimization. Firstly, the cloud constructs a dynamic clustering mechanism and reduces intra-group statistical heterogeneity based on model features and data distribution information uploaded by an edge terminal, and the edge terminal performs local training and intra-group model aggregation according to a cloud clustering result to improve the consistency and adaptability of a local cluster model; secondly, a decoupling knowledge transfer mechanism is adopted, the global model and the local cluster model are decoupled into a feature layer and a classification layer respectively, hierarchical knowledge alignment is carried out in the distillation process, the learning ability of the local model for intermediate feature expression and classification decision boundaries is enhanced, and the distillation efficiency is improved; therefore, the convergence speed and generalization performance of the model in the heterogeneous data environment are improved. Therefore, the technical problems of model performance reduction and weak generalization ability caused by data heterogeneity in the cloud-edge federation in the prior art are solved.

The invention discloses an aviation fleet capacity and airline network matching method and system, and belongs to the technical field of airline operation management. The method comprises the following steps: 1) determining a basic set according to an airline network, and obtaining associated parameters; 2) constructing a two-stage model: firstly generating a transport capacity distribution scene of a multi-group model combination through integer linear programming, then inputting scene data and actual case data as decision units into an SBM-DEA model for solving, calculating a corresponding matching degree, and constructing a virtual boundary and a real boundary of an actual case; calculating the distance between the virtual boundary and the actual boundary of the actual case, and determining the difference between the virtual boundary and the actual boundary so as to determine a matching degree score; and 3) dynamically optimizing fleet configuration according to the matching degree of the aviation fleet capacity and the airline network and the matching degree score obtained by the two-stage optimization model.Real-time quantification of the matching degree is realized through the two-stage optimization model, the hysteresis defect of a traditional static method is overcome, and the utilization efficiency of fleet resources is remarkably improved.

Embodiments of the present application provide a model construction system, method, electronic device and storage medium. A model grouping unit is configured to, in response to a model selection instruction for a model database, obtain a main model and a plurality of candidate models corresponding to the main model, combine the main model with each candidate model to obtain at least two model groups, a feature grouping unit is configured to obtain a first model attribute and a first model feature of the main model in each model group, a second model attribute and a second model feature of the candidate model, and group the first model feature and the second model feature according to the first model attribute and the second model attribute to obtain a corresponding feature set, the feature set including a first feature set and a second feature set, and a model construction unit is configured to fuse the first feature set and the second feature set corresponding to each model group to obtain a fused model corresponding to each model group.

The present invention relates to a multifunctional drone-assisted asynchronous cluster personalized federated learning method, which belongs to the field of wireless communication technology. The method comprises the following steps: S1: establishing a drone-assisted FL network system model; S2: establishing a drone logistics transportation and assisted FL training model; S3: establishing a communication delay, model training time, and uplink delay model between the drone and the sub-server; S4: establishing a personalized federated learning mechanism, solving the personalized model in the inner layer and the global model in the outer layer, and training the personalized model through asynchronous two-layer parallel optimization and intra-group synchronous federated averaging mechanism; S5: optimizing the drone flight path using the group model staleness as the optimization variable; S6: adjusting the optimization variable according to the relationship between the model staleness and the number of user device training times, and proposing a drone path dynamic optimization algorithm based on deep reinforcement learning. The multifunctional drone path optimization algorithm provided by the present invention significantly improves the overall communication efficiency.

The embodiment of the invention provides an account distinguishing method and device, equipment, a storage medium and a program product, and relates to the field of financial science and technology. The method comprises the following steps: acquiring account data, a device fingerprint and a relationship between the account data and the device fingerprint, wherein the device fingerprint is used for identifying a device; generating a credibility evaluation result between the account data and the equipment fingerprint according to the relationship between the account data and the equipment fingerprint by using a pre-constructed credibility evaluation model; constructing a group model according to the account data, the device fingerprints, the relationship between the account data and the device fingerprints and the credibility evaluation result; and based on the group model, according to the credibility evaluation result, determining a group type corresponding to a group to which the account data belongs. According to the method, the group type to which the account belongs is dynamically identified through the group model and the credibility evaluation result, so that the group type to which the account belongs is accurately identified, and the accuracy of determining the group type to which the account belongs is improved.

Feature importance is critical to understanding how predictive models produce accurate results, and can change significantly for different models. The present invention is used to achieve a good ranking for stable feature importance. An optimized technique is presented which considers feature importance value variation within different groups of cross-trained models. Feature importance is computed for all group models with this optimized method, and then a best set of models can be selected based on classification error as well as optimized stable feature importance values.

The present invention provides a traceable federated incremental learning method based on group feature aggregation, which relates to the field of federated learning technology and includes the following steps: S1, performing traceable task learning on end nodes; performing local pruning and fine-tuning based on the initial shared model, and determining whether the task is repeated based on the similarity of label distribution; S2, building a cloud-end collaborative grouping model; sending the global shared knowledge generated by the server model and the aggregation weights corresponding to the global shared knowledge to the end nodes; S3, at the end nodes, the client updates its own model parameters based on the local cross-entropy loss and the global shared knowledge generated by the server model of each group and the aggregation weights corresponding to the global shared knowledge, and outputs the updated model; S4, replacing the new sub-model or updated sub-model in S1 with the updated model, and repeating S1 to S3 until the specified number of training times is reached and training is stopped. The present invention aims to effectively address the challenges brought by storage limitations and task repetitiveness.

The invention provides a virtual power plant element accurate aggregation scheduling method considering demand response, which comprises the following steps: performing aggregation modeling on an energy storage equipment group according to equipment charge and discharge number constraint, charge and discharge power constraint, energy constraint and periodic electric quantity balance constraint, and constructing an aggregation energy storage model; carrying out aggregation modeling on the interruptible load group according to the interruptible interval, the non-interruptible interval and the cuttable time period, and constructing an aggregation interruptible load group model; carrying out aggregation modeling on the translational load group according to the number of translation state equipment and the translational interval, and constructing an aggregation translational load group model; carrying out aggregation modeling on the transferable load group according to the number of the transfer state devices and the total number of the devices, and constructing an aggregation transferable load group model; and constructing a virtual power plant scheduling model according to the above model for solving, and generating a scheduling strategy of the virtual power plant. According to the method, the solving efficiency of the scheduling model is improved, and a more efficient and more reliable virtual power plant scheduling decision is realized.

The invention provides a longitudinal federation joint reasoning method and system based on privacy protection and a storage medium, and the method comprises the steps: an initiating terminal executes local reasoning based on a local target feature data set and reasoning model parameters, and generates an intermediate result; determining the closest classification cluster as a competition cluster, calculating similarity measurement and disclosing a cluster index; the similarity measurement and the intermediate result are processed through secret sharing, share information is generated and exchanged with other terminals; the share information is aggregated according to the public cluster index, and complete similarity measurement and reasoning output of each group of models on the corresponding cluster are obtained through reconstruction; performing weighted fusion on the reasoning ability score by utilizing similarity measurement to obtain an evaluation value, and mapping the evaluation value into a model output weight through a nonlinear amplification function; and finally, performing weighted summation on each group of model output according to the weight to obtain an overall reasoning result. The inference efficiency of the longitudinal federal model under the secret sharing mechanism can be improved.

The invention provides a load group combination regulation capability credibility evaluation method. The method comprises the steps of obtaining multi-dimensional operation data of a user group; the multi-dimensional operation data are analyzed, a load group high-precision model is constructed based on an analysis result, and the load group high-precision model is used for representing dynamic response characteristics of a load group in different scenes; utilizing the load group high-precision model to predict and obtain multi-dimensional load dynamic response characteristics of the user group; and evaluating the overall regulation capability credibility of the load group based on the load dynamic response characteristics, the load declaration value and the actual value to obtain an evaluation result. According to the invention, the comprehensiveness and accuracy of the evaluation result can be improved, and a basis is provided for fine scheduling and decision making of a power grid.

The application discloses a kind of based on function-morphology interaction's building group automatic layout method and system, including the following steps: the design boundary data of target block, road network data, ecological environment data and existing building data are collected, and three-dimensional space digital sand table is constructed;Collect the building area of several case blocks, surrounding road attribute, building storey and land function data, construct land function and space form gray comatose matrix;Factor analysis is carried out to obtain function-morphology factor loading matrix file data, generate function-morphology association model based on knowledge graph;Function-morphology dynamic interaction building group generation is carried out using deep deterministic policy gradient algorithm, and building group layout scheme is generated;Input upper planning, output 3D block building group model object set after screening in accordance with the requirements of upper planning;3D block building group model is superimposed on digital sand table by screening, and the final scheme is determined to be output.

The application discloses a learning resource recommendation method and system based on project field knowledge and user comments, and the method comprises the following steps: collecting learner information, learning resource characteristic information and teacher information, wherein the learner information comprises learner description information and learner interaction information with the learning resource, and the learning resource characteristic information comprises learning resource description information and learning resource characteristic information; finding a teacher with the highest similarity to the learner according to the learner information, and obtaining a matching score of a target learning resource according to the teacher characteristics through a convolutional neural network; establishing a learner short-term preference model and a long-term preference model, and fusing the two models to obtain a learner personal preference model; establishing a learner group preference model, and fusing the learner personal and group models to obtain a learner preference model; and establishing a learning resource characteristic information model and a field knowledge model by using various information characteristics of the learning resource according to the learning resource characteristic information, so that the accuracy of learning resource recommendation is improved.

Disclosed in the present disclosure are a model quantization method, a model operation method, a device, a medium, and a program product. The model quantization method comprises: grouping model parameters of each network layer in a model to be quantized; determining a quantization parameter group of any group on the basis of a target integer bit number required by a network layer to which the any group belongs, and quantifying the model parameters of the any group on the basis of the quantization parameter group; and replacing the corresponding model parameters in said model with the quantized model parameters of each group to obtain a quantized model. The model parameters are grouped for quantization to reduce the influence of outliers present in the model parameters on quantization errors.

The invention provides a wind power and photovoltaic field group resource scheduling optimization method and system, and relates to the technical field of energy scheduling, and the method comprises the steps: constructing a three-dimensional field group model according to a collected equipment set which is a set of all equipment in a wind power and photovoltaic field group; based on the three-dimensional field group model, calling a predetermined clustering strategy to perform clustering analysis on the equipment set to obtain a clustering result; a first node index of first equipment is analyzed, a first initial cluster center is determined, and the first equipment refers to any equipment in a first cluster in the clustering result; performing collaborative analysis on the first resource scheduling record of the first initial cluster center to obtain first prediction information; and carrying out resource scheduling optimization on the wind power and photovoltaic field group based on the first prediction information. According to the invention, the technical problem that the scheduling strategy cannot be adaptively adjusted according to the actual scene in the prior art can be solved, and the technical effect of improving the adaptability of new energy scheduling is achieved.

The application discloses a fusion large model intelligent teaching quality evaluation improvement system and relates to the technical field of wisdom education.The application realizes automatic identification of skill response behavior by means of a disturbance group modeling module and a behavior characteristic difference set of students between original questions and candidate comparison questions.System no longer depends on subjective judgment of teachers, but analyzes answer stability of students when facing semantically equivalent and structurally stable questions through quantitative index.When it is detected that the behavior path of students deviates by a high amplitude, the system can immediately determine that there are skill dependence and concept misplacement problems.For example, after answering the disturbance question, the correct rate of the answer decreases, which indicates that there is a strategic answer instead of concept reasoning, and the objectivity and traceability of the teaching quality evaluation are further improved through the knowledge graph of the original question stem.

The invention discloses an attraction domain boundary determination method and device of a wind power system and computer equipment. The method comprises the following steps: constructing a mathematical equation set model corresponding to a wind power system; constructing a random disturbance model based on the change factors of the wind power system; adjusting the mathematical equation set model based on a random disturbance model to obtain a target mathematical equation set; solving the target mathematical equation set to obtain a phase trajectory diagram; and determining a target attraction domain boundary of the wind power system based on the phase trajectory diagram. The technical problem that the risk of system chaotic oscillation existing in the attraction domain boundary of the chaotic attractor cannot be determined due to unstable output of a wind power system caused by uncertain factors such as wind speed change is solved.

In some implementations, a model system may receive an indication of the set of models that are associated with a set of data points. Each model in the set of models may have been selected using a grid search. The model system may receive, from a user device, a query associated with a selected data point in the set of data points. The selected data point may be associated with a corresponding model in the set of models. The model system may provide information included in the query to the corresponding model in order to receive a result associated with the selected data point. The model system may transmit, to the user device, the result in response to the query.

Provided are a method and apparatus for managing an object, a device, and a medium. The method comprises: respectively acquiring a plurality of responses of a plurality of objects for a media item, wherein the media item comprises a question, and the plurality of responses are respectively encrypted responses of the plurality of objects for the question; determining at least one response type of the plurality of responses; on the basis of the question and the at least one response type of the plurality of responses, dividing the plurality of objects into at least one group, wherein a group among the at least one group corresponds to a response type among the at least one response type; and on the basis of related data of the objects in the at least one group, determining a sample data set used for updating an object grouping model. In this way, users can be divided into groups without knowing specific information in a user response. Furthermore, related data of users in each group can be used for determining a machine learning model, thereby achieving a desired purpose.

The application discloses a momentum-based hierarchical compromise model federated learning method, comprising the following steps: according to the group model and the global model issued by the corresponding edge node, updating the local momentum, the group momentum and the global momentum in the local device, obtaining the first local momentum, the first group momentum and the first global momentum, and training the local model according to the first local momentum; according to the first group momentum and the first global momentum, aggregating the second group momentum and the second global momentum in the edge node, training the group model according to the second group momentum, and aggregating the third global momentum in the center server according to the second global momentum, training the global model according to the third global momentum, and ending the iteration until the global model obtained in the latest iteration is determined to be converged. The application can improve the learning efficiency of the federated learning method.

This invention discloses an EEG emotion recognition method based on hybrid experts and multi-scale spatiotemporal convolution, specifically relating to the field of emotion recognition technology. The method includes the following steps: Step 1: EEG signal preprocessing, acquiring raw EEG signals, and sequentially performing filtering, artifact removal, resampling, and slicing operations. This invention uses a hybrid expert architecture as its core, combined with a multi-scale spatiotemporal convolutional network to construct an EEG emotion recognition scheme. Through group modeling with a "global leader + group experts," it balances individual difference adaptation with feature extraction completeness, improving robustness across scenarios and achieving accurate and efficient recognition of EEG emotions across subjects. Simultaneously, it considers the model's lightweight nature and real-time performance. This invention can effectively analyze data on the transformation states and probabilistic characteristics of traditional Chinese medicine constitution, facilitating understanding of these transformation states and their probabilistic characteristics. Practical application results are good.

The invention provides a construction method of a multi-granularity process knowledge base and a process file configuration type compilation method. The method is based on construction of a process model, a process step model and a process group model. A parameterized and structured work step model is formed through element combinations of different attributes; and then the plurality of process step models are combined to form the modular process model, the process content is complete and high flexibility is achieved through the process step selection units in the process step models, and the modularization degree is improved through the process group model formed by combining the process models. According to the process file configuration type compiling method, direct calling of knowledge modules with different granularities can be achieved, a process route can be rapidly formed, compiling of the process file can be rapidly and accurately completed by configuring different parameters, and meanwhile standardization and consistency of the process files compiled by different technicians can be guaranteed. According to the invention, systematic management and application of process knowledge are realized, and the process compiling efficiency and normalization are improved.

The application provides a pyramid knowledge distillation framework model compression limit analysis method, comprising the following steps: constructing N groups of online deep mutual learning models in a pyramid structure; performing online deep mutual learning on each group of online deep mutual learning models, and recording the parameter quantity and model performance of two models in each group of online deep mutual learning models; wherein, starting from the second group of online deep mutual learning models from bottom to top, while performing online deep mutual learning, the previous group of online deep mutual learning models is accepted for offline knowledge distillation; the potential representation of all models from the first group to the N-1th group is extracted and sent to an adapter to generate teacher importance weight soft labels; the Nth group of online deep mutual learning models is subjected to online deep mutual learning, and the parameter quantity and model performance of the Nth group of models are recorded; and the balance point of the model compression ratio and accuracy is analyzed according to the parameter quantity and model performance of two models in each group of online deep mutual learning models and the parameter quantity and model performance of the Nth group of models.

The invention discloses a model quantification method, a model operation method, equipment, a medium and a program product. The quantification method comprises the following steps: grouping model parameters of each network layer in a to-be-quantified model; determining a quantization parameter group of any group according to a target integer digit required by the network layer to which any group belongs, and quantizing the model parameter of any group according to the quantization parameter group; and replacing the corresponding model parameters in the to-be-quantized model with the quantized model parameters of each group to obtain a quantized model. Model parameters are grouped to be quantized, so that the influence of outliers existing in the model parameters on quantization errors is reduced. When quantization is carried out according to the groups, due to the fact that the quantization parameters in the quantization parameter groups are in a non-linear relation, quantization can be completed on the model parameters in the groups in a non-linear mapping mode according to the quantization parameter groups, and the result generation capacity during model operation is ensured.

The invention discloses a packaging box group support optimization method and system based on nonlinear programming, and relates to the technical field of optimization, and the method comprises the steps: obtaining the real-time environment humidity data of a box body and the real-time weight data of the box body; according to the real-time environment humidity data and the real-time weight data, real-time moisture content data corresponding to the box body is obtained through calculation; calling a preset material mechanical property database, performing mapping matching according to the real-time water content data, and extracting corresponding dynamic pressure-bearing threshold data and dynamic friction coefficient data; constructing a nonlinear programming group support model, setting dynamic pressure-bearing threshold data as a stress constraint boundary stacked in the vertical direction, and setting dynamic friction coefficient data as an anti-slip constraint boundary arranged in the horizontal direction; solving and generating grouping scheme data by using a nonlinear programming grouping model on the premise of satisfying a stress constraint boundary and an anti-slip constraint boundary; and outputting the grouping scheme data to an execution terminal.

The invention discloses a discrete agent-based load group modeling and consistency regulation and control method and system, and the method comprises the steps: building a discrete time load group model based on a load group, and enabling the discrete time load group model to comprise a leader and a plurality of followers; determining a system matrix including an adjacent matrix and a leader connection matrix; dividing the followers by using competition relationship representation; obtaining a low-gain state feedback matrix; setting the maximum response time of regulation and control; in the regulation and control process, a complete intermittent control protocol and a partial intermittent control protocol are generated; respectively calculating control tracks of followers under different control protocols and corresponding saturation constraint values; calculating a consistency error between the system states of the leader and the follower at the regulation and control time k; and judging whether the regulation process is ended or not according to the consistency error. According to the method, control paradigm innovation, heterogeneous system compatibility, time-varying demand adaptation, communication overhead optimization, stable performance guarantee and system self-evolution can be realized, and the flexibility, high efficiency and engineering practicability of power load regulation and control are remarkably improved.

The invention belongs to the technical field of artificial intelligence, and particularly relates to a semi-asynchronous federal learning method and device based on poisoning attack defense and a storage medium, and the method comprises the steps: in any training round, a server receives a first preset number of local models arriving at first; grouping the local models based on the old degrees of the local models; screening out candidate local models from the local models in each group, and carrying out intra-group model aggregation to obtain a group representative model corresponding to each group; determining an equivalent local gradient of each group of representative models based on each group of representative models, a global model obtained in the previous training round and an equivalent global gradient of the global model, estimating the equivalent global gradient of the training round, determining candidate group representative models and inter-group aggregation weights of the candidate group representative models based on the equivalent local gradient, and executing inter-group aggregation to obtain a candidate group representative model; and determining a global model obtained in the training round, thereby realizing the poisoning attack defense of semi-asynchronous federal learning.