263 results about "Dependency information" patented technology

The invention discloses a special reasoning model-oriented data set construction method, and particularly relates to the technical field of data set construction. According to the method, domain task text information, rule constraint information and causal dependency information are analyzed, and a dependency structure chart of a target reasoning task is constructed; on the basis of the dependency structure diagram, topology path expansion and anti-fact topology path backtracking analysis are carried out according to the semantic relation direction and the constraint reverse relation, forward path topology feature data and reverse topology verification data are formed, and therefore a forward and reverse dependency consistency matrix is established for topology conflict recognition, and a consistency correction result is output; and finally, according to a consistency correction result, carrying out topology label labeling, topology equivalence judgment and multi-path admission screening to obtain a topology fidelity sample set, and carrying out dependent signature labeling and topology label coding to generate a special reasoning model training data set of the target reasoning task. And the reasoning path reliability and the data set structuring level of the special reasoning model are improved.

The invention discloses a multi-mode communication protocol optimization method based on deep learning, and the method comprises the following steps: S1, collecting operation data of a communication terminal, and carrying out the preprocessing of the operation data; s2, constructing a multi-mode communication protocol structure diagram, and generating an adjacent matrix; s3, constructing a structured LSTM model, and modeling time sequence characteristics of a communication environment and structural dependence information between protocols; s4, performing performance prediction by using the structured LSTM model; s5, constructing a protocol cause subgraph between the time sequence feature vector and the performance prediction vector; and S6, performing multi-target comprehensive evaluation on the communication protocol, and determining an optimal communication protocol configuration scheme. According to the method, structured deep learning and causal modeling are fused, a multimode communication protocol selection process is optimized, and the method has the advantages of high prediction precision, high decision interpretability and good environmental adaptability.

The invention provides an agent evaluation method and device, an agent, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, in particular to the technical fields of large models, intelligent search, smart questions and answers and the like. The agent evaluation method comprises the steps that according to received evaluation demand information, evaluation task planning is conducted on a specified agent through a large model, an evaluation sample is obtained, the evaluation sample comprises sample demand information and environment dependence information, and the specified agent executes a specified task on the sample demand information according to the environment dependence information, outputting an execution result matched with the sample demand information, wherein the environment dependence information is used for prompting an environment condition on which the specified agent depends on executing the specified task; and evaluating the specified intelligent agent according to the execution result to obtain an evaluation result.

The invention provides a cross-modal feature fusion method based on joint attention, which combines multi-modal data from different sources to predict the relative position of a target to be tracked, and comprises the following steps of: firstly, designing a cross-modal data representation network model for the data of different modals; generating a time sequence high-dimensional implicit code retaining key dependency information from the original data through a code conversion algorithm; secondly, a heterogeneous attention network unit is designed, and features of time sequence high-dimensional implicit codes corresponding to different modal data are extracted through two different forms of attention; and finally, through a cooperative training mechanism, information complementation among different modal data is utilized to generate the most effective feature representation for a time sequence prediction model, and accurate prediction of the moving position of the tracked target is realized.

The invention relates to a task processing method and device of a power grid dispatching cloud platform based on multiple computing resources, computer equipment, a storage medium and a computer program product. The method comprises the following steps: performing task decomposition processing on a to-be-processed calculation task to obtain a plurality of sub-calculation tasks; extracting a first feature vector of calculation complexity of each sub-calculation task, a second feature vector of task data volume, a third feature vector of response timeliness information and a fourth feature vector of hardware dependence information, and performing fusion processing to obtain a fusion feature vector of each sub-calculation task; inputting the fusion feature vector into a trained computing resource prediction model to obtain a target computing resource corresponding to each sub-computing task; and determining a resource load value of the target computing resource, and performing corresponding task processing on the corresponding sub-computing task according to the target computing resource under the condition that the resource load value is smaller than a resource load threshold value. By adopting the method, the processing efficiency of the calculation task can be improved.

The invention discloses a task allocation method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining a to-be-processed task from a global task pool, and analyzing a descriptor of the to-be-processed task to obtain priority information and / or dependency information; determining a processing sequence of the to-be-processed tasks by utilizing the priority information and / or the dependency relationship information; obtaining channel load information from a state register for directly accessing the corresponding channel from the memory; and selecting a target channel from the channels by utilizing the channel load information and combining a load balancing strategy, and distributing the to-be-processed tasks to the target channel according to the processing sequence. According to the task allocation method and device, during task allocation, based on at least one of the priority information and the dependency relationship information and the channel load information, load balancing can be dynamically achieved, at least one of the priority and the dependency relationship of the task can be considered, and therefore the reliability of task processing is guaranteed, and the task allocation efficiency is improved. And different task scheduling requirements can be dynamically met.

The invention relates to a method, a device and equipment for identifying ancient characters of a war bamboo slip, and a medium. The method comprises the following steps: acquiring a to-be-identified war bamboo slip image containing a plurality of ancient characters; flattening a local feature graph containing local detail information of ancient characters in a to-be-recognized war national bamboo slip image output by the OfficientNetV2 convolutional neural network into a sequence structure, and splicing the sequence structure with a feature sequence containing global semantic information and long-range dependency information of the ancient characters output by a Transform encoder in a channel dimension to form a joint feature tensor; layer normalization processing is carried out; and inputting the combined feature tensor subjected to layer normalization processing into a classification module of an ancient character recognition model so as to output a single character recognition result corresponding to a single ancient character or character sequence recognition results corresponding to a plurality of continuous ancient characters. Compared with the prior art, the method has better supplementing and repairing capacity when processing defective and broken strokes, and the recognition capacity for characters with incomplete structures or complex forms is remarkably improved.

The invention discloses a power robot abnormal target detection method based on improved YOLOX. According to the method, image data shot by an electric power robot are acquired, an improved YOLOX model which introduces a multi-scale cross-level local network MS-CSPNet and a small target decoupling detection head is used for processing an image, and accurate detection and positioning of an abnormal target are realized. The MS-CSPNet enhances the feature expression ability of a small target through multi-path convolution, a decoupling detection head is combined with deep convolution, expansion convolution and 1 * 1 convolution, context and long-range dependency information is effectively extracted, classification and regression task separation is achieved, and the detection precision is improved. Each power robot model is continuously optimized based on a federated online incremental learning algorithm, multi-terminal cooperative training is supported, adaptive updating of the model is realized while data privacy is guaranteed, and the method is widely applied to power equipment state monitoring and fault early warning scenes.

A computer-implemented system analyzes program and data flows in a software system comprising code written in multiple programming languages using a generative large language model (LLM) directed by programming-language-specific prompts. The LLM identifies functional components within the code, generating labeled graph nodes that include a node type, a node name, and dependency information. A graph construction computer system processes the labeled graph nodes to generate a directed graph, where nodes represent functional components and directed edges represent dependencies. The system stores the graph in a database and provides a web-based interface for visualization, allowing users to explore, query, and analyze program and data flows across the software system. The system enables automated, language-agnostic dependency mapping, facilitating software analysis, debugging, and modernization.

The invention provides a test program generation method, electronic equipment and a storage medium, relates to the field of chips, and is mainly technically characterized by comprising the following steps: generating a plurality of subprograms corresponding to each module in at least two modules according to instruction characteristics of the at least two modules in a target processor; wherein each subprogram comprises dependency information with other subprograms, each subprogram comprises a plurality of instruction fragments, each instruction fragment comprises at least one instruction, and the dependency information of the subprogram is used for indicating whether the instruction fragment in the subprogram conflicts with the instruction fragments in other subprograms or not; the to-be-used instruction fragments are continuously selected until a preset ending condition is met, and the next to-be-used instruction fragment in every two adjacent to-be-used instruction fragments is the to-be-used instruction fragment which is sequentially selected from any subprogram of any module and does not conflict with the previous to-be-used instruction fragment; and obtaining a target test program according to a selection result of the to-be-used instruction fragment.

This invention discloses a method and system for hyperspectral image denoising based on a spatial-spectral joint self-attention mechanism, belonging to the field of image processing technology. First, based on the characteristics of hyperspectral images, a spatial-spectral joint self-attention mechanism network is constructed. The noisy hyperspectral image is used as input to the network to extract spatial-spectral features. Then, a global spectral self-attention mechanism is used to extract the band correlations of the hyperspectral image. Finally, the extracted spatial-spectral features are reconstructed using a multiple perceptron and residual connections to reconstruct a clean, noise-free hyperspectral image. The system includes a feature extraction subsystem, a non-local spatial self-attention subsystem, a global spectral self-attention subsystem, and an image reconstruction subsystem. This invention can effectively restore noisy hyperspectral images to obtain noise-free hyperspectral images. Compared to convolutional networks, it can better model long-range dependency information and has better adaptability to target hyperspectral images.

The invention relates to a load prediction method TSR-MAEN based on decomposition representation and an attention encoder network. Firstly, the algorithm provides a trend-seasonal decomposition representation method based on comparative learning so as to enhance the decoupling performance of a model on a trend component and a seasonal component of a load sequence and the representation capability of decomposition characteristics. Secondly, in order to deeply mine and decompose time sequence information contained in a feature hidden space and dependency between features, the algorithm proposes a mixed attention encoder network; through segmented embedding, a trend-season dynamic projection mixed self-attention mechanism and a multi-head mixed output mechanism, trend and season multi-dimensional features after load sequence decomposition are effectively utilized to express and enhance the long-range dependence information extraction capability of the model. Experiments show that the TSR-MAEN algorithm provided by the invention has excellent performance in an electrical load prediction task of oil and gas development, and the prediction error is reduced by 8.51% compared with that of an optimal baseline algorithm.

The invention discloses a software outsourcing-oriented heterogeneous code warehouse intelligent coupling degree detection system and method, and relates to the technical field of software engineering.The method comprises the steps that an instruction is detected, and code file information, dependency information and development log information are obtained through a distributed crawler module, an abstract syntax tree analysis tool and a natural language processing module; carrying out modular segmentation on the code file by utilizing semantic coupling analysis, inputting the code file into a large language model equipped with a LoRA adapter, identifying a coupling inverse mode, and carrying out preliminary optimization based on a quantum heuristic coupling optimization algorithm; after optimization, a multi-user concurrent scene is simulated through digital twin simulation, and the coupling degree is dynamically detected; meanwhile, a cross-language interface testing tool is used for testing interface compatibility and call chain coupling; the system comprises an information acquisition module, a semantic coupling analysis module, a quantum optimization module and a coupling degree detection module, provides a deep optimization basis for developers, and improves the coupling degree detection and optimization efficiency of the heterogeneous code warehouse.

Apparatuses and methods for a CSI report configuration for CSI predictions in one or more domains. A method performed by a user equipment (UE) includes transmitting capability information indicating capability of the UE to support machine learning (ML) based channel state information (CSI) prediction in one or more domains, receiving configuration information that indicates parameters for ML based CSI prediction in the one or more domains; receiving CSI reference signals (RSs), and measuring the CSI-RSs. The method further includes determining, based on the configuration information and the measured CSI-RSs, a plurality of CSI predictions in the one or more domains; determining a CSI report including one or more of the plurality of CSI predictions and dependency information indicating dependencies between CSI predictions in the plurality of CSI predictions; and transmitting the CSI report.

The invention relates to the technical field of data processing and the field of insurance businesses and smart medical treatment, and provides a code review method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: obtaining code change information; performing context-aware retrieval processing on the code change information based on the code multi-dimensional knowledge base to obtain code context dependence information; checking the code change information and the code context dependence information based on a code checking module to obtain initial code comment information; screening the initial code comment information based on a preset value filtering module to obtain screened code comment information; performing adoption evaluation processing on the screened code comment information to obtain code correction evaluation information; and performing optimization processing on the code knowledge graph, the code checking module and the value filtering module according to the code correction evaluation information. Through the technical scheme, the code can be intelligently reviewed and processed, and the code review efficiency is improved.

The invention relates to a cross-view positioning method and system based on hierarchical gating fusion and sequence space aggregation, and belongs to the field of computer vision. The method comprises the following steps: extracting three different levels of feature maps from a pre-trained visual model as the input of a gating residual fusion module; a final fusion feature map is obtained through a gating residual fusion module; and finally fusing the feature map to obtain a feature vector through a spatial displacement and selective state space joint aggregation module. According to the method, the problem that the overall information expression is insufficient due to the fact that feature fusion lacks content self-adaption and interlayer guidance and feature aggregation cannot keep local geometry and long-range dependency information at the same time is solved.

The invention discloses a code review method and device, electronic equipment and a storage medium. Comprising the steps that a full-amount source code file of a target project is obtained from a code repository according to a code change event, multiple programming languages and dependency information corresponding to the programming languages are determined based on the full-amount source code file, and each source code file in full-amount source codes corresponds to one programming language; determining at least one to-be-reviewed source code file in the plurality of source code files, and extracting annotation information in each to-be-reviewed source code file; determining a cue word template corresponding to the to-be-reviewed source code file, and determining cue words of the to-be-reviewed source code file based on the programming language, the dependency information, the annotation information and the cue word template corresponding to the to-be-reviewed source code file; and processing each to-be-reviewed source code file and the corresponding cue word based on a preset large language model to obtain a code review result corresponding to each to-be-reviewed source code file. Flexible construction of the cue word is realized, and the accuracy of code review can be improved.

The invention provides a test case generation method and device, electronic equipment and a readable storage medium, code change is extracted based on Git change perception, and code change information of a to-be-tested program of a target version is obtained; screening out a first test case based on the code change information and a trained test case screening model; generating a second test case based on the code change information and a trained test case generation model; extracting dependency information of each external dependency associated with the code change, and matching a Mock object and a Mock rule for the external dependencies in the first test case and the second test case; and determining the matched first test case and second test case as a target test case of the to-be-tested program of the target version. Thus, through the trained test case screening model and test case generation model, the test case matched with the code change of the to-be-tested program can be generated, and the accuracy of test case generation in automatic testing is improved.

Embodiments of the present disclosure relate to a request processing method and apparatus, a device, and a storage medium. The method provided herein comprises: receiving a target request to be processed; during the processing of the target request, determining a triggered first component; in response to a first expected duration of the first component not exceeding an available duration of the target request, determining first dependency information of the first component, wherein the available duration is determined on the basis of a preset duration and an executed duration of the target request, and the first dependency information indicates a dependency type of a second component to be executed on the first component; and at least on the basis of the first dependency information, determining a target execution strategy for the first component, wherein the target execution strategy indicates whether to skip the first component.

The invention discloses a cross-language interaction malicious behavior detection method based on a high-order function, and relates to the technical field of computer security. The method comprises the following steps: receiving a source code of a certain program, modeling a cross-language interface in the program, and identifying and marking a cross-language calling boundary; different language source codes are converted into universal intermediate representations (IR), the IR is combined, and cross-language calling points are marked; constructing a global cross-language function call graph and a control flow graph, determining a function analysis sequence by inverse topological sorting to perform dependency analysis on each function, and abstracting dependency information in the processes into a function abstract in a high-order function form; and parameter instantiation is carried out when function calling is processed, so that the dependency relationship between the cross-language APIs is obtained. According to the method, efficient analysis and transmission of the cross-language dependency relationship are realized through the high-order function abstract and the dependency matrix, the high overhead of constructing a complex system dependency graph by a traditional method is avoided, and the analysis efficiency and the detection accuracy are improved.

A computer-implemented system analyzes program and data flows in a software system comprising code written in multiple programming languages using a generative large language model (LLM) directed by programming-language-specific prompts. The LLM identifies functional components within the code, generating labeled graph nodes that include a node type, a node name, and dependency information. A graph construction computer system processes the labeled graph nodes to generate a directed graph, where nodes represent functional components and directed edges represent dependencies. The system stores the graph in a database and provides a web-based interface for visualization, allowing users to explore, query, and analyze program and data flows across the software system. The system enables automated, language-agnostic dependency mapping, facilitating software analysis, debugging, and modernization.

The invention provides a big data platform configuration risk propagation detection and analysis method based on LLM, and the method comprises the steps: carrying out the semantic analysis of the natural language description of a vulnerability problem through LLM, and carrying out the standardized replacement of a complete field according to a risk knowledge base, and obtaining risk elements; performing document analysis on the multiple target configuration resources, and performing semantic-level matching analysis on the structured document fields by utilizing LLM and according to the risk elements to obtain a risk anchor point set; according to the component dependency information and the risk anchor point set, sequentially performing path construction, path context information packaging, natural language query prompt construction and path propagation judgment by using LLM to obtain a risk propagation path set; and sequentially carrying out map construction and map format standardization on the risk propagation path set to obtain a risk propagation map. Therefore, the accuracy of risk propagation detection is high, the flexibility is good, and the influence path of the configuration risk in the big data platform is easier to assess.

A method for generating a bill of materials file includes a generation device obtaining target information, where the target information includes training dependency information, model composition information, and model metadata. The training dependency information is information about a training resource for training an artificial intelligence (AI) model. The model composition information is information about an intermediate model in a process of training the AI model, and the model metadata is attribute information of the AI model. The generation device generates a bill of materials file of the AI model, where the bill of materials file includes the target information.

The embodiments of the present disclosure provide a container unit scheduling method, a server, a program product and a storage medium. The container unit scheduling method comprises: in response to a scheduling request for a first container unit, acquiring service dependency information of the first container unit; on the basis of the service dependency information, determining from among a plurality of candidate computing nodes a target candidate computing node for deploying the first container unit; and after the target computing node is determined, creating the first container unit on the target candidate computing node, wherein the service dependency information is acquired on the basis of traffic monitoring data of a first workload in the first container unit over historical periods, thereby enabling feedback on the actual dependency of the first workload on other services. Thus, using service dependency information as the basis for scheduling container units to select candidate nodes can facilitate scheduling of container units having a dependency relationship to the same computing node, so that it is not necessary for different container units having a dependency relationship to perform communication interaction across nodes.

The application discloses an improved short-term and nowcasting precipitation prediction method using random masks and a Transformer, and belongs to the field of precipitation prediction. The improved short-term and nowcasting precipitation prediction method using random masks and a Transformer comprises the following steps: S1, constructing a random mask spatiotemporal sequence image; S2, constructing a network model, and inputting the spatiotemporal sequence image marked with the mask into the network for model training; the network model comprises an encoder-decoder structure with a UNet as a core model, a SwinTransformer module is embedded in the encoder, and an SE-Net attention mechanism is introduced; S3, in the model training process, a prediction value is obtained through a forward propagation process of the input image, then the model is continuously fine-tuned according to a loss function, the loss function is minimized, and the accurate prediction capability of the model is realized; and S4, L1+L2 regularization is used in the training process to prevent overfitting. The high-order non-stationarity in the modeling spatiotemporal sequence is improved, and the short-term and long-term dependence information in the spatiotemporal sequence is learned at the same time, so that the prediction accuracy of the model is improved.

Systems and methods for deploying dependent updates include mechanisms for requiring that dependency information be provided for each update so that dependent updates may be identified. Update dependencies are tracked so that dependent updates are not deployed until parent updates have been completed. Deployment sequencing is implemented on top of existing asynchronous deployment policies so that asynchronous workflow remains intact and unaltered. The asynchronous workflow is upgraded to a synchronous (i.e., sequential) workflow for updates having dependencies to ensure that updates having dependencies are applied in the correct order.

A method processes data requests from a requester to a responder over a network by using remote direct memory access. The method includes the requester issuing two or more requests into a work queue of a network interface controller. The method further includes the network interface controller executing the queued requests by sending them to the responder, and receiving responses to the sent request from the responder. One or more given requests each includes dependency information related to the response from the responder to one or more previous requests in the work queue, the method further includes, before executing the given request by the network interface controller, evaluating the dependency information and executing the given request depending on the result of the evaluation.

The invention discloses a kernel and driver upgrading method of an operating system, and relates to the technical field of computers, and the kernel and driver upgrading method comprises the following steps: loading a plurality of target drivers of a target operating system in a target kernel environment to determine a plurality of loading results corresponding to the plurality of target drivers; obtaining a driver priority from a preset driver knowledge base, and screening out a to-be-upgraded driver from the plurality of target drivers according to the driver priority and the plurality of loading results; obtaining target dependency information of the to-be-upgraded driver from a preset driver knowledge base, and compiling and testing the to-be-upgraded driver based on the target dependency information and the driver priority to obtain a target driver package of the to-be-upgraded driver; and completing upgrading of the kernel and the driver of the target operating system based on the target kernel environment and the target driver package. Therefore, screening and grading mechanisms can be added after kernel compiling is completed, the drive upgrading range is effectively narrowed, uncertainty caused by drive upgrading is reduced, and customer service stability is improved.

The invention discloses an ESB intelligent message routing optimization method based on reinforcement learning, and the method comprises the steps: deploying a Kafka ESB node, registering node information, and carrying out the heartbeat synchronization maintenance of a node state; the node registers an HTTP service, updates the state change, and synchronizes the service state between cities; summarizing node, service, message and dependency information, and constructing a four-dimensional routing state tensor; receiving a service request, generating a unique identifier, and executing state recombination and dependency recoding; inputting the comprehensive evaluation into a decision maker, and selecting an optimal routing path by utilizing MuZero prediction; and after the service is executed, analyzing the feedback factor, generating a feedback vector, and updating a routing enhancement strategy parameter. According to the method, intelligent optimization and continuous adaptive adjustment of ESB message routing in a cross-city asynchronous environment are realized by constructing a four-dimensional routing state tensor and a reinforcement learning decision-making mechanism based on MuZero.

The present disclosure relates to a software project dependency package management method, an electronic device and a computer readable storage medium. The software project dependency package management method comprises: detecting a security vulnerability of a dependency package in a software project based on a security vulnerability database and a dependency information file; generating a repair scheme according to the detection result; and updating the dependency information file to repair the security vulnerability according to the repair scheme, wherein the dependency information file comprises information for building a dependency tree of the software project.