24 results about "Structural consistency" patented technology

The invention discloses a reasoning request processing method and electronic equipment, and relates to the technical field of model reasoning, and the method comprises the steps: generating a topology execution graph based on a calculation flow captured after initialization or deployment of a pre-training language model, and constructing a graph pool containing exclusive topology execution graphs of different sequence lengths in combination with the distribution characteristics of historical sequence lengths, the construction logic of the graph pool can cover various sequence length scenes appearing at high frequency in practical application; meanwhile, based on an execution graph construction mode captured by a pre-training language model calculation process, the structural consistency and calculation correctness of topological execution graphs with different lengths are ensured, so that the reasoning framework can flexibly respond to reasoning requests with different lengths; when the actual sequence length of the reasoning request is greater than the preset length threshold, the reasoning request is divided into a plurality of independent sub-blocks, then for each sub-block, the corresponding target topology execution graph is matched from the graph pool and the execution is started, and the situation that the super-long sequence goes back to the dynamic kernel function due to the fact that the super-long sequence exceeds a fixed range is avoided.

This invention discloses an open-domain adaptive method based on inter-class relationship modeling and unknown sample mining, comprising the following steps: acquiring source and target domain datasets; performing supervised pre-training in the source domain; retaining the feature extractor and constructing an extended classifier with virtual class weights; extracting target domain sample features; constructing a class relationship distribution and structural consistency metric; and completing initial pseudo-label assignment; introducing inter-class separation and intra-class compactness constraints; optimizing the structural space reservation loss; and constructing a feature discrimination buffer region; constructing a high-confidence unknown negative sample set based on prediction unknowns and structural deviations through dual threshold screening and stability verification; optimizing the boundary reinforcement loss using the negative sample set; and integrating multiple loss functions for iterative training until model convergence. This invention solves the technical problems of existing open-domain adaptive methods, such as difficulty in balancing the cross-domain transfer effect of known classes with the recognition accuracy of unknown classes, susceptibility to negative transfer, and degradation of the discriminative structure in the feature space.

The application relates to the technical field of network security and software testing, and discloses a generative fuzzy testing method for a PDF reader, which has the technical effects of reducing the cost of syntax modeling, improving the coverage of the model, and effectively suppressing the hallucination of a large language model by designing a polymorphic object syntax reduction template and automatically constructing a POG reduction library by combining a retrieval enhancement generation technology; wherein the POG template encodes the complex semantics of the PDF specification into a machine-interpretable structured skeleton by introducing inheritance, polymorphism and dynamic generation instruction fields; the RAG technology combines the specification text, the table and the prior knowledge of the large model to automatically extract accurate object definitions from the ISO standard document; and finally, the POG reduction library with high coverage and high fidelity is constructed by combining the breadth-first iterative extraction and manual checking, thereby fundamentally solving the problems of high cost of traditional manual modeling, poor structural consistency of the deep learning method and hallucination caused by direct generation of the large language model.

The invention belongs to the technical field of geologic modeling, and particularly discloses an interactive geologic model construction method, which remarkably improves the efficiency, flexibility and accuracy of geologic model construction by introducing an intelligent modeling auxiliary technology, and supports geologic experts to guide model generation in a visual operation mode. According to the method, rapid definition and dynamic adjustment of key structures such as faults and horizon are automatically carried out, and a more efficient and credible modeling process is realized, so that the rationality, the expression ability and the structure consistency of a geologic model are greatly improved. According to the method, the problems of'black box 'operation, difficulty in modification, high subjectivity, difficulty in processing complex structures, uncertainty expression and the like in traditional geological modeling are solved, a perceptible, guidable and verifiable man-machine interaction geological modeling system is constructed, the controllability, adjustability and traceability of the texture modeling process are realized, and the method is suitable for mass production. Rapid iteration and interdisciplinary integration of the texture model are facilitated, and the method has remarkable technical advantages and application value.

This invention discloses a PCB defect detection method based on spatial geometric continuity constraints. This method introduces spatial geometric continuity regularization loss into the deep detection model, explicitly constraining the spatial geometric continuity of similar defects during training, thereby reducing fragmented predictions and improving detection coherence and robustness. This invention overcomes the limitations of existing methods that only focus on single-target detection accuracy. It innovatively introduces a spatial geometric continuity constraint mechanism at the loss function level, and combines it with industrial scenario data augmentation strategies, a single-stage target detection structure, dynamic learning rate scheduling, and a detection evaluation mechanism. This transforms the detection results from "independent discrete predictions" to "holistic predictions with spatial structural consistency," thus significantly improving the spatial coherence, stability, and engineering interpretability of the detection results while ensuring detection accuracy, ultimately enhancing the practical value and reliability of industrial defect detection systems.

The invention discloses a multi-modal image matching method based on phase consistency structure information distillation. According to the method, on a to-be-matched multi-modal image pair, firstly, stable structure prior information is obtained through phase consistency calculation, and the stable structure prior information and multi-scale features extracted by a convolutional neural network jointly form structure-guided feature expression; performing structural distillation training on the network by using structural consistency loss and cross-modal consistency loss, so that the obtained coarse-scale and fine-scale features have higher structural consistency under a cross-modal condition; on the basis, coarse-scale global matching is completed by adopting a self-attention and cross-attention alternately stacked feature transformation module to generate candidate matching pairs, and then local refinement is performed on the candidate matching pairs through fine-scale feature matching to obtain a final matching set. According to the method, the phase consistency structure prior and the deep learning matching framework are deeply fused, and the matching precision and the calculation efficiency are improved while the generalization of cross-modal matching is ensured.

The invention provides a large model enhancement training method based on an industry knowledge graph, and the method comprises the steps: constructing an industry task structure graph, carrying out the structure coding of task nodes, and obtaining a task structure vector set and a task structure hop count matrix; performing stage division on the training tasks based on the atlas to form a stage task set and a corresponding stage sample set; the large model is trained stage by stage, structural consistency constraint is introduced in each stage, so that internal task representation of the model is aligned with a task structure vector, and the relationship between tasks conforms to a hop count matrix; and finally, fusing the model parameters of each stage and the task intermediate representation, optimizing a unified structure space to enable the task representation of the final model to be close to an original map structure, retaining stage training memory and maintaining a hop count distance relationship between tasks. According to the method, the structural understanding and generalization ability of the large model in the industry task is effectively improved.

The application discloses a kind of water turbine unit fault mechanism diagnosis methods based on knowledge graph structure enhancement, first, construct the water turbine unit fault mechanism knowledge graph covering multidimension;Then the fault description input to be diagnosed is input, and the BM25 model is used to carry out semantic recall, obtains candidate fault set;Then, for each candidate, calculate its structural consistency score GraphSim with fault description in four dimensions, and based on score, the candidate is rearranged, and the highest score candidate is output as the final diagnosis result.The mechanism link of final diagnosis can also be input into local large language model, and natural language explanation is generated.The application is self-consistent by structural consistency rearrangement, and the diagnosis result is forced on complete fault mechanism chain, which significantly improves the diagnosis accuracy, interpretability and robustness, and is simple to implement, and can be widely popularized to various complex industrial equipment intelligent operation and maintenance scene.

The invention relates to the field of knowledge graph updating and representation, and provides a continuous embedding learning-based electric power knowledge graph updating method, which comprises the following steps of: performing semantic coding on a newly-added triple and a historical triple in an electric power knowledge graph snapshot, and extracting deep semantic representation of the newly-added triple and the historical triple; based on the candidate similar triple set, calculating the semantic similarity between the target entity or relationship and historical knowledge, distributing a normalized weight, and generating an initial embedding representation of a newly added entity or relationship through weighted average; a mask automatic encoder is utilized to reconstruct a local sub-graph of a target node, and the structural consistency of new and old knowledge is kept; on the basis, an embedding learning optimization strategy based on a TransE model is adopted, and meanwhile, a lightweight regularization mechanism is introduced to inhibit historical embedding excessive offset. According to the method, the problem of disastrous forgetting can be effectively relieved when the electric power knowledge graph is dynamically updated, the consistency of new knowledge and old knowledge in a semantic space is ensured, the reasoning and prediction capabilities of the knowledge graph are improved, and the method has relatively high stability.

The invention discloses a geostatistical modeling method uncertainty quantification method based on a mutual information entropy framework, and relates to the technical field of geostatistical modeling quality control, and the method comprises the steps: collecting the space sample data of an estimation domain, determining target methods, constructing a parameter matrix for each target method, and carrying out the calculation of the parameter matrix; carrying out multiple modeling on each target method according to the parameter matrix on the estimation domain to generate a plurality of modeling results; dispersing each realization modeling result to a unified grid to obtain a realization modeling result set of each grid, counting distribution of a plurality of realization modeling results for each grid, constructing probability distribution of each grid, calculating point entropy of each grid based on the probability distribution, and forming a spatial uncertainty entropy field. According to the method, quantification of spatial uncertainty is achieved by achieving probability distribution and entropy measurement, IMUI indexes are constructed through joint entropy, conditional entropy and mutual information, and evaluation of output convergence and structural consistency of the method is achieved.

This invention belongs to the field of medical data processing technology and discloses a joint prediction method, system, storage medium, and electronic device for medical expenditures based on structural consistency constraints. Addressing the inherent defect of inconsistent prediction head output logic in existing multi-task learning methods, this invention forcibly eliminates semantic conflicts between different prediction heads during the training phase by simultaneously applying non-cross-quantile constraints, nested risk probability monotonic constraints, and cross-head consistency regularization. The measured quantile cross-rate and risk monotonic violation rate are both 0%, achieving deployment-level structural consistency guarantees and solving the technical problem of insufficient reliability in automated decision-making systems caused by contradictory output logic in existing technologies.

The invention relates to a multifunctional bitwise operation method and system based on double-array co-location index, in particular to the field of communication, and remarkably improves the reliability and efficiency of the operation system by introducing mode and data collaborative verification and structural consistency deep analysis. In the initialization stage, the compatibility of the operation mode and the input data is subjected to prospective judgment, operation errors and resource waste caused by mismatching of the data and the mode in a traditional method are effectively avoided, the system can intelligently analyze the operation requirement, structural consistency verification is conducted on the multi-source input data, and the system reliability is improved. On the basis of ensuring that subsequent operation is established on a reliable data basis, an optimized bitwise operation rule is adopted in an execution stage, calculation precision and consistency are guaranteed, final output is synthesized and optimized, the result is accurate, the structure is normative, the whole scheme realizes full-process optimization from data verification, operation execution to result output, and the calculation efficiency is improved. And the overall performance and stability of the system in processing complex and changeable scenes are greatly improved.

The present application relates to the field of computer information processing, and particularly relates to a multi-source heterogeneous knowledge graph construction method and system. The method comprises the following steps: accessing ancient books, clinical literature, and data of traditional Chinese medicine, drugs and disease databases, and forming an entity candidate set after field mapping extraction; calculating the name similarity, attribute similarity and semantic vector similarity of the entity candidate set, and inputting the multiple types of similarity into a fusion network to generate an entity alignment mapping table; constructing a remote supervision annotation set and training a relationship classification model, driving expert annotation incremental update by model uncertainty, and outputting a relationship triple set and a corresponding relationship confidence set; finally forming a three-layer heterogeneous graph structure connected by syndrome type, disease, prescription and drug components, and writing into a graph database to generate a knowledge graph database. The present application realizes unified modeling of entity alignment, relationship extraction and relationship strength quantification in multi-source heterogeneous data, and improves the structural consistency of the knowledge graph construction process.

The invention discloses an incomplete alignment multi-view clustering method based on structural consistency comparative learning, and belongs to the technical field of computer vision. The method comprises semantic consistency comparative learning and view consistency comparative learning, the semantic consistency comparative learning realizes implicit class alignment of samples by capturing global semantic consistency information, and the view consistency comparative learning learns consistency information between views while retaining local structure information of the views by hierarchically selecting positive samples. Therefore, an accurate sample corresponding relation is established.

The invention discloses a citation network node classification method and system based on a hidden equidistant manifold. The invention provides a node classification algorithm for a citation network structure. The method comprises the following steps of: firstly, maximizing mutual information between paired sub-graph views related to a literature citation relationship through comparative learning by adopting a path mask strategy so as to fully capture structural characteristics of a citation network; meanwhile, calculating the matching degree between the degree prediction of the paper node and the degree of the original node in the mask quotation graph so as to ensure the structural consistency; and obtaining a potential space coordinate keeping the equidistant constraint through equidistant learning. The method can effectively improve the classification accuracy and robustness in the citation network node classification task, not only solves the node classification problem of high-dimensional complex data in the citation network, but also can retain the manifold structure information of the hidden space of the citation network.

The invention discloses a panel furniture automatic assembly path planning method based on three-dimensional point cloud registration. The method comprises the following steps: constructing an attribute topological graph for a computer aided design model and scene point cloud; structural consistency evaluation is carried out on the multiple initial candidate poses through a graph matching algorithm, so that ambiguity is eliminated, and a unique optimal pose is screened out; performing high-weight optimization on the functional structure by adopting a constraint-aware weighted iterative nearest point algorithm to obtain a final assembly pose meeting assembly constraints; through a path planning algorithm fused with dynamic bias sampling, guidance is carried out in a constraint subspace, and a collision-free motion path conforming to the process is efficiently generated. According to the method, the identification problem of similar plates is effectively solved, the assembly precision is ensured, and the efficiency and feasibility of path planning are improved.

The invention discloses a multi-physical field joint inversion method based on Dual-ResNet structure constraint, and belongs to the technical field of geophysical exploration inversion, and the method comprises the steps: collecting and obtaining multi-source physical data in a to-be-detected region; constructing initial distribution of a velocity model and a resistivity model, and obtaining an input tensor of a velocity field and a three-channel tensor of logarithmic resistivity; constructing two independent residual neural networks, inputting the initial velocity model and the resistivity model into the residual neural networks in a one-to-one correspondence manner, and extracting deep structure features; deep structure features reflecting underground medium space distribution rules are extracted; introducing a total variation regular term and a structural consistency regular term, and constructing a joint objective function; in the joint objective function, taking multi-source physical data as observation data; and iteratively optimizing the speed model and the resistivity model by adopting an alternate updating strategy until a preset iteration convergence condition is reached. According to the scheme, the method has the advantages of being simple in logic, high in applicability, accurate, reliable and the like.

The application discloses a kind of based on index dependency graph constraint's big model generation control method and system for teaching evaluation report generation, belong to the field of teaching, including: constructing teaching evaluation index set, and establish the index dependency graph of the logical dependency relationship between index;In the big model inference stage, generate candidate evaluation text and extract the semantic vector of each index;According to index dependency graph, the consistency constraint calculation and conflict detection are carried out to candidate text, obtain consistency loss and conflict penalty;Based on the above loss, candidate text is scored and reordered, and the optimal text is selected;If conflict exceeds the preset threshold, trigger closed-loop rewriting mechanism to regenerate until the final teaching evaluation report that output satisfies structural consistency is obtained.The application can explicitly introduce index structure constraint in the process of teaching evaluation report generation, effectively suppress cross-index semantic conflict, improve the logical consistency, automation level and explainability of evaluation report.

The application provides an industry knowledge graph-based large model enhancement training method, which comprises the following steps: constructing an industry task structure graph, structurally encoding task nodes to obtain a task structure vector set and a task structure hop number matrix; dividing a training task into stages based on the graph to form a stage task set and a corresponding stage sample set; training a large model stage by stage, and introducing a structural consistency constraint in each stage to align the model internal task representation with the task structure vector and make the relationship between tasks comply with the hop number matrix; and finally, fusing the model parameters and task intermediate representations of each stage, optimizing in a unified structure space, making the task representation of the final model close to the original graph structure, retaining the stage training memory and maintaining the hop number distance relationship between tasks. The method effectively improves the structural understanding and generalization ability of the large model in the industry task.

The invention relates to the technical field of model construction, is suitable for the financial field, and provides a sample expansion method, computer equipment and a computer readable storage medium, and the method comprises the steps: obtaining an expanded sample based on the information type of each sample content in a real sample; calculating the semantic similarity between the expansion sample embedding vector and a vector in a real data embedding index corresponding to the real sample; determining a structural consistency score based on the structured data in the expanded sample; determining semantic stability based on the output embedded representation of the extended sample in the plurality of models; determining the credibility of the expanded sample based on the semantic similarity, the structural consistency score, the semantic stability and the output result confidence of the current initial model; and if the credibility is greater than or equal to a preset credibility threshold, determining that the expanded sample is an effective sample. According to the scheme, training data is constructed for rare class services in a targeted manner, and more diversified reliable training samples are provided for a downstream service model.

This invention discloses a financial statement summary generation system based on artificial intelligence-generated content, belonging to the field of artificial intelligence technology. The system performs segmentation and extraction of cited fragments from the financial statement text, constructs a closed-loop structure by combining candidate fragments from footnotes, and classifies the closure status through correspondence comparison. It then performs backfilling expansion and constraint reconstruction on the closed structure to generate summary generation primitives containing the main body of the matter, the main body of the explanation, and the main body of the limitation. Finally, it verifies the structural consistency and semantic integrity of the summary generation primitives through checks on the clarity of the matter, the correspondence of the explanation, and the completeness of the limitation. This system, focusing on the construction and verification process of the closed-loop structure, structurally reconstructs the cross-segment relationships between the main body of the financial statement and the footnotes, ensuring that the summary content has a clear subject focus, consistent sources of explanation, and complete limiting conditions. It is suitable for scenarios involving automatic financial statement summary generation and assisted review.

This invention discloses a metadata-driven method and system for constructing a large language model intelligent agent through bootstrapping, relating to the field of artificial intelligence technology. The method includes the following steps: collecting natural language requirements generated from business scenarios and organizing these requirements into a unified expression format; generating readable metadata description content within this unified expression format; performing field splitting and role attribution around the metadata description content; constructing data type definition metadata within the unified expression format; and embedding prompt words into the data type definition metadata to generate content. This invention achieves progressive generation from natural language to data types, rules, and intelligent agent definitions through unified metadata expression, supporting dynamic construction and automatic loading at runtime, reducing manual intervention, and improving automation. Simultaneously, new metadata is generated during runtime and fed back into the generation chain, forming a continuous bootstrapping closed loop, enhancing structural consistency and version evolution controllability.

The application discloses a wind farm cluster power prediction method of structural consistency generated graph network, comprising: wind farm multi-source data input and variable definition; decoupling coding of endogenous power sequence and exogenous meteorological variables; wind farm operation mode memory bank construction and sample level mode activation; mode driven wind turbine dynamic graph structure generation; future power rough prediction based on the generated model; structural consistency constraint of the prediction sequence and the real sequence; operation mode consistency constraint; power representation propagation and prediction refinement based on the dynamic graph; joint optimization target; model reasoning and actual deployment. Through the collaborative design of the mode memory bank, the dynamic graph generation and the structural consistency constraint, the application solves the problems in the traditional wind power prediction method, such as the difficulty of the static graph structure to adapt to complex wind conditions, the information interference caused by the multi-source variable fusion, the lack of structural rationality of the prediction result and the like, and can significantly improve the accuracy, the robustness and the engineering applicability of the wind farm cluster power prediction.

The invention is suitable for the technical field of medical image analysis, and provides a medical image edge detection method based on a structural consistency modeling mechanism, and the method comprises the steps: obtaining a to-be-detected medical image; image edge detection is carried out on a to-be-detected medical image by using the trained edge detection model to obtain an edge image of the to-be-detected medical image, and a loss function in the training process of the edge detection model is constructed based on learnable local affine disturbance and structure alignment; the edge detection model comprises a Unet encoder, a structure saliency map construction module, a feature compression module, a structure reconstruction and jump fusion module based on structure consistency guidance, and a Unet decoder. According to the invention, the cooperative constraint of the structure consistency and the topological stability can be realized, and the medical image edge detection accuracy is improved.