18 results about "Domain level" patented technology

The invention provides a cross-domain equipment fault diagnosis method and system based on large and small model cooperation, and relates to the technical field of equipment fault diagnosis. According to the method, the causal field generalization structure is introduced into the small model, explicit decomposition is carried out on the stable causal law and the field specific difference, and meanwhile, the causal field generalization structure is corrected by using the large model, so that the small model can automatically identify and retain the causal relationship which is universally applicable to each device and each field; therefore, the influence of inter-domain distribution difference is effectively eliminated. Theoretical analysis shows that the generalization error of the model mainly depends on the accuracy of the stable causal item, and the structure can minimize error drift caused by distribution drift. Therefore, the robustness of health state evaluation and fault prediction can be remarkably improved in a cross-domain scene, and the fault diagnosis model can still keep the prediction capability close to the training domain level under the condition of no target domain annotation data.

The invention discloses a dynamic core binding and energy efficiency optimization method and device based on a scheduling domain. The method comprises the following steps: analyzing a CPU topological structure when a kernel is started, and identifying and registering a Cluster scheduling domain; constructing a multi-level scheduling domain hierarchy comprising a Cluster layer, an MC layer and an NUMA layer; regularly detecting the utilization rate of the current scheduling domain, and dynamically switching the levels of the scheduling domain according to a comparison result of the utilization rate and a preset threshold value; and when it is detected that the scheduling domain level switching frequency exceeds a preset frequency threshold, starting an oscillation suppression mechanism. Through the three-layer elastic scheduling domain architecture comprising the Cluster layer, the MC layer and the NUMA layer and the dynamic oscillation suppression mechanism, the task scheduling efficiency and the system energy efficiency are optimized, fine-grained load balancing can be achieved, memory access delay can be reduced, scheduling domain level oscillation caused by short-term fluctuation and floating can be avoided, and the system stability can be enhanced.

The invention discloses a wetland remote sensing data cross-domain classification method based on neural network adversarial learning, and belongs to the technical field of image processing, and the method comprises the following steps: S1, collecting cross-domain wetland remote sensing data, extracting spatial features and spectral features through a spatial-spectral feature extraction network, and processing the spatial features and the spectral features to obtain final features; s2, completing domain-level distribution alignment; s3, class level distribution alignment is completed; and S4, outputting a prediction result. According to the method, the adaptive optimization multi-classifier is designed, the adaptive optimization multi-classifier comprises two high-density classifiers and one low-density classifier, prediction is carried out by using the low-density classifier instead of using the high-density classifiers during prediction, and thus the prediction precision of the model in the target domain can be further improved.

A system generates a test interface for display at a client device. The test interface includes an input field for matching criteria. The system receives input of matching criteria from an administrator of a domain associated with the client device. The system searches for files within a content management system repository corresponding to the domain having content that matches the matching criteria. The system updates the representations as the matching criteria is edited. The system receives user input to form a rule based on the matching criteria and monitors for files satisfying the rule. The monitoring may result in a remediation action for files that satisfy the rule. In response to determining that a file that satisfies a quarantine rule, the system updates a permissions data structure to provide and revoke direct permissions to access the file. The system moves the file to the quarantine storage repository.

A system groups multiple entities in a large distributed data store (DDS), such as directories and files, into a subset called a domain. The domain is treated as a unit for defining policies to detect and treat sensitive data. Sensitive data can be defined by enterprise or industry. Treatment of sensitive data may include quarantining, masking, and encrypting, of the data or the entity containing the data. Data in a domain can be copied as a unit, with or without the same structure, and with transformations such as masking or encryption, into parts of the same DDS or to a different DDS. Domains can be the unit of access control for organizations, and assigned tags useful for identifying their purpose, ownership, location, or other characteristics. Policies and operations, assigned at the domain level, may vary from domain to domain, but within a domain are uniform, except for specific exclusions.

This invention discloses a data processing method, apparatus, device, and storage medium. The method includes: acquiring at least two event data sequences generated by a target user in at least two business domains; extracting features from each event data sequence corresponding to each business domain to determine event feature information corresponding to each business domain, and performing information fusion to determine first fused feature information; performing cross-attention processing on pairwise event data belonging to different business domains to determine event-related feature information corresponding to each event data combination, and performing information fusion to determine second fused feature information; and determining target fused feature information based on the first and second fused feature information. Through the technical solution of this invention, coarse-grained fusion at the business domain level and fine-grained fusion at the event level can be combined, improving the data fusion effect between multiple business domains and thus improving the accuracy of user analysis.

The application provides a selective adversarial enhancement network and system for rolling bearing fault diagnosis, a framework of the selective adversarial enhancement network is integrated with a balance enhancement module and a selective adversarial module, and is used for realizing fine-grained sub-domain alignment between a source domain and a target domain; the selective adversarial enhancement network comprises the balance enhancement module, the selective adversarial module and an uncertainty suppression module; the balance enhancement module is used for dynamically adjusting category distribution of the source domain and the target domain, so that the category distribution of the source domain and the target domain is balanced; the selective adversarial module is used for screening abnormal categories and accurately aligning shared categories, and realizes distribution alignment at a sub-domain level; and the uncertainty suppression module is used for suppressing uncertainty propagation caused by misclassification by optimizing a loss function, and improving alignment quality of samples close to a decision boundary.

The present application provides a software-based classifier built on a machine learning model that distinguishes between two kinds of malicious URL hosting apex domains: public and private. This classification helps security professionals specify which domain levels to block, the whole apex domain in the case of private apexes or specific subdomains in the case of public ones. The classifier is also built on a machine learning model that differentiates attacker-owned hosting domains from compromised hosting domains. This distinction is crucial to help security operators take the appropriate mitigation actions. For example, attacker-owned domains could be blocked permanently whereas compromised ones temporarily.

The application relates to a scheduling method, device and storage medium of a parallel program, which comprises the following steps: S101, monitoring a process running in a first-level scheduling domain of a processor system, and giving a preset value to the process according to the number of sub-processes created by the process in a preset period when the process runs; S102, comparing the preset value with the number of cores in the processor system, if the preset value is greater than or equal to the number of cores, then according to the scheduling domain where the cores loaded by the process are located, increasing the scheduling domain level of the sub-process from the first-level scheduling domain to a second-level scheduling domain by one, and making the sub-process run in the cores within the range of the second-level scheduling domain. The application improves the utilization of processor resources while ensuring the interaction efficiency of the parallel program.

The invention discloses a whole vehicle multi-domain hierarchical pure hardware safety island architecture and a management and control method, and belongs to the technical field of vehicle function safety. The architecture adopts a hierarchical pure hardware design, and comprises a whole vehicle level safety root, a multi-domain level safety island, full-module independent power supply timing, hardware real-time monitoring of heartbeat, power supply and clock faults, and hierarchical hardware execution of fault handling; the whole process of the method is directly detected and controlled by hardware without software judgment. The method is free of firmware and processor, low in fault detection delay and fast in processing response, solves the problems that existing software safety monitoring is prone to failure and slow in response, meets the vehicle function safety use requirements, and is suitable for safety monitoring of all types of vehicles.

The application belongs to the technical field of fault diagnosis, and provides a complex equipment non-steady state condition fault self-adaptive diagnosis method and system. In order to solve the problems that the existing transfer learning method faces in the non-steady state condition, such as coarse global distribution alignment granularity, difficulty in adapting to multi-stage dynamic drift and the like, a fine-grained domain adaptation mechanism based on local maximum mean difference is designed, a graph Laplacian spectral distance constraint is constructed to maintain the topological structure consistency of the cross-domain feature dependency relationship, and the collaborative optimization of the source domain and the target domain in the sub-domain level is realized, so as to realize high-precision and high-robust cross-domain fault diagnosis under the non-steady state conditions such as time-varying speed.

The application relates to the field of natural language processing and the field of deep learning, in particular to a domain difference self-learning fine-tuning text classification method based on a pre-training model, which comprises the following steps: obtaining a text to be measured; adopting a fine-tuned pre-training model to perform feature extraction on the text to be measured, so as to obtain a feature vector of the text to be measured; inputting the feature vector of the text to be measured into a domain level head module, so as to obtain the confidence of the text to be measured in each category; and taking the category with the highest confidence of the text to be measured in all categories as the final prediction result of the text to be measured. The application is based on a pre-training model and combines a domain difference self-learning technology, introduces supervised contrast learning, classification learning and weak clustering learning and other strategies, realizes the learning of different levels of semantic differences, and utilizes the implicit relationship between categories to assist classification. Meanwhile, a classifier weight initialization method more suitable for text classification is adopted, so that the classification performance is improved.

The invention relates to a parallel program scheduling method and device and a storage medium, and the scheduling method comprises the following steps: S101, monitoring a process running in a first-stage scheduling domain in a processor system, and endowing the process with a pre-designed numerical value according to the number of sub-processes created in a preset period when the process runs; s102, comparing the pre-designed numerical value with the number of cores in the processor system, and if the pre-designed numerical value is greater than or equal to the number of the cores, according to a scheduling domain where the cores loading the process are located, scheduling the process according to the scheduling domain where the cores loading the process are located; and increasing the scheduling domain hierarchy according to which the sub-process is created by one level from the first-level scheduling domain to a second-level scheduling domain, so that the sub-process runs in the core within the range of the second-level scheduling domain. According to the method, the utilization rate of processor resources is improved while the interaction efficiency of the parallel programs is ensured.

The invention discloses an Internet of Vehicles decentralized cross-domain identity authentication method based on a block chain under strict supervision, and relates to the field of data security, and the method comprises the steps of system initialization, domain-level initialization, roadside base station and vehicle registration, vehicle pre-authentication, cross-domain communication signature, signature verification and vehicle revocation. According to the method, a decentralized trust system is constructed by adopting a block chain alliance chain and a decentralized identity technology, so that the dependence on a centralized certificate mechanism is thoroughly eliminated; through a selective disclosure mechanism based on a Merkle grid, minimum disclosure and fine-grained privacy control of authentication attributes are realized; and a deformation signature algorithm is designed, and a hidden label which can be decrypted only by a specified receiver is embedded in a standard signature, so that the sensitive content cannot be recovered even if a supervision mechanism obtains all private keys of the system, and the vehicle privacy is effectively guaranteed under strict supervision. The method has the advantages of attack resistance, low delay and low communication overhead, and is suitable for a large-scale dynamic car networking environment.

The invention discloses a training method and device of a power grid load prediction model and a storage medium, and the method comprises the steps: dividing a plurality of calculation power nodes into a current round cooperation domain, and determining a current round domain head calculation power node and a current round domain member calculation power node in the current round cooperation domain; determining the domain-level pruning rate of the round based on the resource condition information of the round; scheduling each current-round domain member computing power node to perform model training on a local model based on the current-round domain-level pruning rate and the current-round initial load prediction model parameters to obtain current-round model parameters; scheduling a target current round domain head computing power node to aggregate and upload the current round model parameters; carrying out cross-domain aggregation processing on the uploaded edge model parameters of the round; if the cross-domain aggregation parameter of the round does not reach the training termination condition, carrying out model training of the next round through a member computing power node of the next round domain; and if the cross-domain aggregation parameter of the round reaches a training termination condition, performing the last round of model training through each domain member computing power node according to the cross-domain aggregation parameter of the round.

System and methods for providing stand-in services at a domain can include obtaining a service request at a domain, determining one or more services at the domain to fulfill the service request, in response to determining a service of the one or more services is unavailable, providing a model as a stand-in service for the service, determining, by the one or more services and the model, a decision based on the service request, and sending, in response to the service request, the decision as output by the domain level. The model can provide the stand-in service by obtaining data for the service based on the service request context, identifying one or more keys based on the obtained context data, retrieving, based on the one or more keys, data from a cache, and applying the data to the model, the decision being based on the data applied to the model.

Techniques described herein relate to a method for managing a distributed multi-tiered computing (DMC) environment. The method includes normalizing, by a local controller associated with an DMC domain, task resource demand dimensions for each task associated with a scheduling job; summing the resource demand dimension for each task to generate resource demand vectors; classifying tasks based on priority; sorting tasks based on associated resource demand vectors; obtaining critical path, earliest start time, and latest start time associated with each task; sorting tasks based on critical path and earliest start time; and generating scheduling assignments based on the priority, capacity of devices in a final candidate list, resource demand vectors, earliest start time, and the critical path.

The application discloses a kind of hierarchical topology domain weight-aware task scheduling method and system, the method includes: based on the network bandwidth difference between node performance and inter-node modeling, obtain the hierarchical cluster node performance topology domain model with weight, can reflect the network bandwidth difference between node performance and inter-node.Resource scheduling process, task specifies topology domain level, scheduler traverses all nodes of each topology domain under this level and carries out two stages of preselection and optimization, finally selects a best node, and the node is distributed to the node, and the scheduling result is recorded;If multiple topology domains meet the requirements, the node resources preferred by the Pod distributed in each topology domain are added, and the binding request is sent according to the highest score of the topology domain scheduling result, and the specific binding action is responsible by kubelet.The application considers the network bandwidth difference between node performance and inter-node, can schedule workload to the best performance domain, and improve the efficiency of large model training.