22 results about "Distributed knowledge" patented technology

The invention discloses a large language model auxiliary optimization strategic decision-making system and a large language model auxiliary optimization strategic decision-making method. The system comprises six core modules. The dynamic knowledge fusion module constructs a three-layer distributed knowledge network, constructs an entity association weight matrix through a bidirectional Transform model based on an attention mechanism, and realizes knowledge dynamic association in combination with a time attenuation factor and a hybrid coding technology. The large language model module performs field fine tuning by adopting incremental pre-training and low-rank adaptation technologies, and introduces an exclusive word segmentation list to improve professional analysis precision. The full-process intelligent writing module covers submodules for report generation, revision and the like, and supports full-life-cycle management of reports. The strategic decision intelligent deduction module integrates scene impact factors, and realizes multi-scene deduction through reinforcement learning and Monte Carlo tree search. The interaction display module provides a visual interface, and the multi-mode interaction module realizes full task chain management. According to the invention, real-time knowledge support and intelligent deduction capability are provided for strategic decision making, and decision making efficiency and accuracy are improved.

The invention relates to the technical field of software architecture design, and discloses a software architecture design method based on intelligent bionic implementation of a large model group. The method comprises the steps of obtaining software design requirement input; performing multi-level semantic analysis on software design requirement input to generate an initial component set and context environment description; based on the context environment description, candidate component knowledge related to the initial component set is retrieved from a distributed knowledge base; performing consistency verification on the candidate component knowledge and the initial component set by utilizing a dynamic inference engine to generate a verified component map; according to the verified component atlas, candidate software architectures are generated through an architecture template matching algorithm; and performing iterative adjustment on the candidate software architecture by adopting an optimization strategy, and outputting a target software architecture design. Through collaborative analysis and intelligent reasoning of the large model group, automation and intelligence of software architecture design are achieved, and the efficiency and quality of software design are improved.

The invention discloses a multi-modal document data processing method and system for large language model training, and the method comprises the steps: receiving a plurality of original documents in various formats, extracting the structure information of each original document, and recognizing a text region and an image region of each original document based on the structure information; performing optical character recognition on the text region and the image region by adopting a parallel OCR (Optical Character Recognition) engine based on GPU (Graphics Processing Unit) acceleration and heterogeneous calculation to generate recognition text data of the corresponding original document; performing multi-dimensional quality evaluation and cleaning on the recognition text data of each original document, and outputting normalized text data; and storing the standardized text data into a distributed knowledge base according to a predefined structure, and performing copyright and compliance test on the standardized text data. By adopting a parallel OCR recognition engine based on GPU acceleration and heterogeneous calculation, efficient and high-precision batch processing of multi-modal documents is realized, and the processing speed, the recognition accuracy and the data quality are improved.

The invention discloses a federal forgetting method and system based on projection gradient rise, and belongs to the technical field of machine learning, and the method comprises the steps: obtaining information of N clients and a central server in a federal, and carrying out the T-round training of the federal through a federal weighted aggregation learning algorithm; performing data distillation on each client to obtain a distillation data set; acquiring a federal global model and a local model of the forgotten client in T-1 rounds, and solving a reference model on the forgotten client; the client optimizes parameters in an l2 norm ball with the radius of delta around the reference model; sending the federal global model to all clients, and performing opposite training strategies on the forgotten clients and the remaining clients; re-aggregating the local models of the forgotten client and the remaining clients; and obtaining a final federal forgetting model until E rounds of training are completed. The problems that a non-distributed machine forgetting method caused by distributed knowledge penetration is difficult to be used for federal learning and disastrous forgetting is caused by gradient rise of federal forgetting are solved.

A ship pipe fitting recognition system based on multi-source data enhancement and intelligent recommendation comprises a multi-source data enhancement module, a model training module, a recommendation decision module, a federated learning platform and a block chain evidence storage module, the system is of a'cloud + side 'double-layer architecture, the cloud is used for intelligent evolution, distributed knowledge is integrated through federated learning, and the block chain evidence storage module is used for storing the distributed knowledge. The global model is continuously updated and optimized, the edge end is used for edge real-time response, and efficient pipe fitting feature detection and pipe fitting number recommendation are carried out at the equipment end. According to the method, the detection precision and the non-standard pipe fitting recognition capability in complex illumination and shielding scenes can be remarkably improved, and the misjudgment rate is reduced; the decision-making problem of similar part hybrid scenes is solved, and the assembly efficiency and accuracy are improved; meanwhile, the data privacy security is ensured; the technical blank in the aspects of data diversity, algorithm robustness and industrial adaptability in the field of ship pipe fitting recognition is filled, the efficient requirement of a ship assembly scene is met, and technical support is provided for ship manufacturing intelligence.

The invention provides a set of cross-agent collaborative system which deeply fuses a DIKWP five-layer semantic structure (data, information, knowledge, wisdom and intention) and a block chain. The system performs semantic annotation on each transaction on a chain by using an ontology / knowledge graph to form a traceable multilayer semantic account book; semantic registration and query of a heterogeneous agent are supported through a standardized interface; an intelligent contract with semantic recognition and reasoning capabilities is introduced to automatically trigger cooperation, intention matching and right settlement; and low-level data is continuously improved into high-level knowledge and insight by using a synchronization mechanism of on-chain / off-chain collaborative reasoning and event driving. Through encryption, access control and governance mechanisms, the platform realizes semantic interoperation, credible sharing and intention alignment, the block chain is expanded from data evidence storage to a distributed knowledge base, the method is suitable for multi-party cooperation scenes such as medical consultation, supply chains, smart cities and scientific research, and performance and expansibility (such as side chains, subscription and under-chain reasoning) are considered at the same time.

The invention provides a safety reasoning method and device based on multi-modal data, equipment and a storage medium, and the method comprises the steps: carrying out the logic data collection of a to-be-monitored region, and obtaining a first data block; performing data transmission and knowledge enhancement on the first data block according to a preset logic strategy library to obtain a second data block; performing distributed feature extraction and logic fusion on the second data block to obtain a fused feature data set; performing hybrid reasoning according to the fused feature data set and a preset distributed knowledge base to obtain a risk assessment result; and generating an executable strategy set according to a risk assessment result.

This invention provides a comprehensive platform that integrates 3D virtual tours with a multi-agent Al-driven system for enhanced user engagement, scalability, and autonomous management. The platform includes a 3D Virtual Tour Generator that captures and renders physical spaces into interactive, high-resolution 3D environments. A network of specialized Al agents - including a Platform Guide Agent, Customer-Specific Agents, Visitor Assistant Agents, Compliance Agents, and Engagement and Marketing Agents - drives the platform's interactive capabilities. These agents are powered by distributed knowledge bases, enabling them to provide context-aware assistance, real-time responses, and proactive user guidance. The platform supports self-management, allowing businesses to create and control their digital presence, including configuring Customer-Specific Agents. Additionally, agent-to-agent interactions facilitate continuous learning, platform maintenance, and improvement. The invention is designed for various applications, including real estate, education, tourism, and retail, offering a human-centric, scalable solution that enhances virtual tours and provides autonomous, high-quality information 24 / 7, reducing reliance on human guides.

The present invention discloses a multi-stage task processing method and system based on an intelligent agent model, relating to the field of task processing technology. The method comprises collecting heterogeneous data streams through a distributed sensor network and generating dynamic feature vectors using a quantum annealing-inspired feature encoder; calculating the expected utility value of paths using a Bayesian optimization algorithm and projecting the high-level task space into a Koopman space; initiating multi-threaded asynchronous computing, collecting execution status data in real time, and constructing a causal graph model; integrating short-term execution logs through a neural Turing machine, calling edge computing nodes for distributed knowledge extraction, generating multimodal interpretable reports, and updating long-term memory. By initiating multi-threaded asynchronous computing, collecting execution status data in real time, constructing a causal graph model, and outputting an execution result matrix with confidence scores, the present invention improves the stability and reliability of task execution.

The invention discloses a distributed knowledge fusion and vector retrieval system and an implementation method thereof. The distributed knowledge fusion and vector retrieval system comprises a local node, an encryption transmission component and a federated server. The local node is configured with a multi-mode encoder to generate an Embedding vector; the encryption transmission component encrypts the data; and the federated server receives the encrypted vectors to construct a federated vector knowledge base, and configures an active generation neural cache module. The module analyzes continuous encrypted query vectors through a built-in time sequence prediction model, predicts a future query intention, and actively manages cache according to a prediction result. According to the method, through active predictive caching and generative concept understanding, the problems that in the prior art, sudden query response is slow and emerging concept retrieval is difficult are effectively solved, meanwhile, federal learning and homomorphic encryption technologies are utilized to guarantee whole-course data privacy security, and the response performance and knowledge discovery ability of the system are remarkably improved.

The invention relates to a distributed knowledge base file storage system based on a high-speed optical fiber network. A hardware architecture comprises storage nodes, an optical fiber switch and an application server. The system optimizes communication through a remote direct memory access technology, and a software architecture comprises a high-speed network driving layer, a joint file system daemon process and a user space file system framework. The united file system daemon process provides a unified virtual directory view, and supports standard file operation and data consistency. And the network optimization synchronization module optimizes data transmission by using a zero copy technology and asynchronous I / O. And the dynamic configuration management module supports dynamic node management and persistent configuration. And the control management interface module provides a UNIX Socket interface, supports JSON format instructions and realizes dynamic scheduling of storage resources. The user space file system framework is responsible for daemon process management, provides a unified storage view service and supports large model knowledge base application. The system provides an efficient and extensible storage solution for a large-scale knowledge base.

The invention provides an intelligent semantic matching distributed knowledge base query method, apparatus and device. The method comprises the steps of obtaining a query request input by a user; preprocessing the query request to obtain target query data; processing the target query data to obtain key information, intention types and retrieval strategies; according to the target query data, performing similarity query in a vector knowledge base to obtain a preliminary screening result; the vector knowledge base is updated according to a preset rule; according to the key information and the intention type, performing correlation judgment on the preliminary screening result to obtain an intermediate screening result; and filtering and sorting the intermediate screening result according to the retrieval strategy to obtain a target screening result. According to the method, the timeliness, the matching precision, the resource utilization efficiency and the expandability of knowledge query can be improved.

The application relates to the technical field of power system automation and Internet of Things control systems, and particularly discloses an Internet of Things monitoring and intelligent early warning system for power equipment. The system comprises an edge intelligent terminal deployed on site, a regional aggregation server and a cloud coordinator. The edge terminal is responsible for data acquisition, local diagnosis and incremental learning; the aggregation server fuses regional knowledge through federated knowledge distillation; and the cloud coordinator performs global monitoring and update arbitration. Through the federated knowledge distillation framework, distributed knowledge fusion is realized without the need for centralized raw data. The edge terminal only uploads model parameters or outputs soft labels, thereby protecting the privacy and security of power equipment operation data and solving the problem of data compliance use. The teacher model refined by the regional aggregation server can fuse diagnosis experience under diversified working conditions, thereby improving the generalization ability and early warning accuracy of the model for different types of power equipment and operating environments.

Embodiments of the present specification relate to a distributed knowledge graph fragmentation method and device, the method is applied to a distributed system, the distributed system includes a master device and N working devices, the method comprises: any working device clusters nodes in an edge full set assigned to edges in the device subset into a plurality of node clusters, forms a clustering result and sends it to the master device; the master device combines the received N clustering results and sends them to the N working devices; any working device assigns a plurality of node clusters in the received clustering result to a plurality of fragments according to a preset cost evaluation function, forms a fragmentation proposal and sends it to the master device; the master device receives N fragmentation proposals, and determines the first fragmentation result with the minimum cost value as the first target fragmentation result for fragmenting the target knowledge graph.

The disclosed technology generally relates to identification parsing in a distributed knowledge graph. In one example of the techniques, a graph query is received. The metagraph includes identifier acceptance information associated with which of the data storage acceptance identifiers of the plurality of data stores, and identifier derivation information associated with which of the data storage derivation identifiers of the plurality of data stores. The meta-graph and cost information are used to select from among query paths a query path capable of implementing the graph query based on minimizing cost according to a cost metric. Upon determining that a failure has occurred in the selected query path, the selected query path is changed using the metagraph and the cost information. A response to the graph query is provided based on the selected query path.

The application is suitable for the field of rail transit technology, and provides a whole-process AI intelligent agent management and control system based on an integrated wheel set maintenance intelligent workshop, which comprises a multi-source fusion intelligent agent module, a role perception decision intelligent agent module, a device conflict arbitration intelligent agent module and a closed-loop optimization intelligent agent module; the multi-source fusion intelligent agent module is used for collecting the state and process data of equipment, realizing cross-modal alignment through an AI model to generate a feature vector, and updating a distributed knowledge graph based on a time series graph convolution network; the role perception decision intelligent agent module is used for analyzing user permission labels and instructions, and generating a scheduling scheme in combination with equipment load; the device conflict arbitration intelligent agent module is used for monitoring the spatial topology of AGV and mechanical hands; and the closed-loop optimization intelligent agent module is used for analyzing wheel set press-fitting torque data, extracting historical cases of the knowledge graph, and generating process adjustment instructions. The application realizes the whole-process intelligent leap from a data fusion intelligent agent, a dynamic decision intelligent agent to a process closed loop.

A computer system and method for dynamic interaction between human operators and persistent cognitive machines is disclosed. The invention enables adaptive collaboration in operational environments by integrating multimodal translation, cognitive processing, load balancing, trust calibration, operational learning, and team coordination. Human inputs such as voice, gestures, biometric signals, and contextual data are converted into prompts for a cognitive core that processes reasoning through multi-stage language models and thought caching. Operator cognitive load is quantified by combining physiological and behavioral indicators, and tasks are dynamically allocated between human and machine based on load, task complexity, and trust. Operational modes transition between advisory, collaborative, autonomous, and override states with safeguards to ensure stability and human primacy. Outputs are adapted in detail, modality, and timing according to operator state. Continuous learning captures interaction patterns and team dynamics, providing personalized adaptations, distributed knowledge sharing, and resilience to component failures.

The application discloses a general artificial intelligence knowledge infrastructure and a multi-modal knowledge processing method, and belongs to the field of artificial intelligence software systems. In view of the problems of existing multi-modal data semantic fragmentation, fragile knowledge storage format and static and rigid knowledge graph, the application provides a full-stack solution through a four-layer software architecture: a virtual unified knowledge bus (V-UKB) realizes automatic discovery and millisecond to microsecond time synchronization of multi-modal devices; a multi-modal knowledge record format (MKRF) realizes atomic-level fusion of perception data and semantic knowledge, and forms self-description, cross-platform and traceable knowledge units; and a Qian Kun knowledge core realizes neural-symbol hybrid reasoning and dynamic evolution of knowledge. All technical solutions of the application are realized based on general computing devices and open source technologies, do not require special hardware support, and can be widely applied to scenes such as intelligent libraries, long-term archives, cross-modal intelligent retrieval, distributed knowledge services and the like.

The invention is suitable for the technical field of rail transit, and provides a whole-process AI agent management and control system based on an integrated wheel set maintenance intelligent workshop, and the system comprises a multi-source fusion agent module which is used for collecting the state and process data of equipment, achieving the cross-modal alignment through an AI model, generating a feature vector, and storing the feature vector into a database; the distributed knowledge graph is updated based on the time sequence diagram convolutional network; the role perception decision-making agent module is used for analyzing the user permission label and the instruction and generating a scheduling scheme in combination with an equipment load; the equipment conflict arbitration agent module is used for monitoring the spatial topology of the AGV and the manipulator; and the closed-loop optimization agent module is used for analyzing the wheel set press-fitting torque data and extracting historical cases of the knowledge graph to generate a process adjustment instruction. According to the method, the whole process from a data fusion agent and a dynamic decision agent to a process closed loop is intelligently increased.

This invention discloses a reconfigurable reaction flywheel torque control system, comprising a data acquisition unit, a model building unit, a flywheel evaluation unit, a trajectory prediction unit, and a torque control unit. It acquires operational data of the reaction flywheel and redundant flywheels to construct a DIKW (Distributed Knowledge Graph) model of the flywheels. The DIKW model is used to store, identify, and transform the data, accurately obtaining the usage status of each reaction flywheel. This allows for the evaluation of which reaction flywheels and redundant flywheels need replacement. The optimal replacement node is then determined based on the spacecraft's trajectory. Replacing the reaction flywheels and redundant flywheels at the optimal node minimizes interference with the spacecraft's attitude. After replacing the reaction flywheels, their usage can be scheduled to be consistent, preventing damage from excessive use of one reaction flywheel from affecting the entire spacecraft. Finally, the torque of the reaction flywheels is controlled to ensure the normal operation of the spacecraft.

The embodiment of the invention provides a federated learning method for non-independent identically distributed data, and belongs to the technical field of distributed machine learning. Comprising the steps that a client executes local model training on a private data set, and a client is randomly selected as a leader student; the super network receives the embedded vector extracted by each client network layer, calculates a feature importance score and executes top-k feature selection; the server collects the selected features, performs layer-by-layer alignment to solve the framework difference of the client model, and performs fusion to generate a comprehensive feature map; a leader student receives the fused feature map and incorporates aggregated knowledge into a model thereof through a self-distillation mechanism. According to the method, the feature with the most information content is intelligently selected through the super network, unified representation of distributed knowledge is achieved in combination with self-distillation, model performance degradation caused by non-independent identically distributed data is effectively relieved, and the method is suitable for federal learning optimization in data heterogeneity scenes such as medical treatment and the Internet of Things.

The invention discloses a distributed knowledge graph construction and encrypted storage method based on a large model, and aims to solve the problems of uncontrollable knowledge graph construction quality, low automation degree and insufficient distributed storage security in the prior art. According to the method, firstly, an overall construction task is decomposed and allocated to a plurality of large model intelligent agents with different capability advantages for distributed knowledge extraction; then, through a multi-round cross validation and consensus achievement mechanism, mutual verification and arbitration are carried out on the extracted preliminary knowledge fragments, the illusion of a large model is effectively inhibited, and high-credibility knowledge is screened out; and finally, performing fine-grained dynamic encryption storage on the high-quality sub-atlas generated by fusion, namely, dynamically generating a key according to the security levels of the entities and the relationships and the structural context of the entities and the relationships in the atlas, performing management in combination with an isomorphic key atlas, and finally performing distributed storage on encrypted data blocks.