95 results about "Model quality" patented technology

The invention discloses a software requirement modeling method and device based on a natural language, equipment and a medium, and relates to the field of software requirement modeling. According to the software demand modeling method and device based on the natural language, the equipment and the medium, through automatic demand analysis and model generation, the manual workload is greatly reduced, and meanwhile, errors caused by manual understanding deviation are reduced; close connection among the steps is ensured by sharing intermediate representation and a unified data model, output of the previous step provides structured information for input of the next step, and information loss or distortion is avoided; a demand provider is allowed to participate in a model confirmation process through an iterative demand acquisition and modeling framework of man-machine cooperation, and the quality of a demand model is continuously improved; by means of the model consistency analysis technology based on the data flow dependency relationship, the problem of inconsistency between the requirement specification and the software model can be automatically detected, and the software product quality is improved.

The invention relates to the technical field of large model algorithms, in particular to a large model dynamic compression optimization method and system based on sparse pruning, and the method comprises the steps: capturing original weight fluctuation data generated by resource fluctuation in reasoning, and obtaining sparse weight reference data through sparse processing; analyzing calculation complexity through model reasoning delay data, and separating reasoning delay amount caused by a model scale; dynamically controlling the model compression ratio within a preset performance range based on the delay amount and the sparse reference data, and collecting reasoning precision distribution data under different compression parameters; evaluating a model performance state under each parameter by means of a neural network simulation method, and generating a performance state simulation result; determining a model quality optimization compensation parameter based on a simulation result by combining resource fluctuation data acquired in real time in a compression process; and finally, the compression strategy is adaptively regulated and controlled through the compensation parameters, and collaborative optimization of model calculation complexity, reasoning precision and delay during dynamic change of hardware resources is realized.

The invention provides a model quality evaluation optimization method and device for industrial simulation application and a server, and relates to the technical field of simulation application generalization development, and the method comprises the steps: obtaining a to-be-processed geometric model, and carrying out the format detection processing, format conversion processing and asynchronous processing of the geometric model, and determining a target input model; performing grid division processing and finite element analysis processing on the target input model, determining a finite element calculation result, performing local error estimation analysis processing on the finite element calculation result, locally refining a model grid of the target input model, determining a model key region, performing local encryption processing on the key region, and obtaining a local encryption result of the target input model; determining a model output result; and carrying out optimization processing on the model output result through a model quality evaluation optimization system with a layered architecture, and determining an optimized target model output result. The model quality can be remarkably improved.

A privacy protection heterogeneous federal learning method with Byzantine robustness includes training a local model and calculating a sketch, removing an abnormal local model, selecting a client with quick response capability, encrypting local update and submitting ciphertext, executing weighted aggregation and distributing an updated global model. The method has the beneficial effects that a high-dimensional local model is converted into a low-dimensional sketch by adopting locality sensitive hashing, and the quality of the local model is effectively evaluated on the premise of protecting data privacy. Then, the Byzantine clients are identified and removed based on the hierarchical clustering technology, aggregation weights are distributed to the remaining clients according to the local model quality, and the global model convergence speed is increased; in addition, by selecting the clients which are high in response speed and have representative data sets to participate in model training, the problem of outdated persons caused by system isomerism is solved under the condition that the global model accuracy is not affected.

The invention relates to a dynamic multi-dimensional quality evaluation system and method for a thermal power plant intelligent inventory supervision model. Real-time data flow of the DCS is collected and compared with the model prediction value, and a prediction error is generated; based on the load rate, the fuel characteristic and other characteristics, the current operation state is matched to a preset working condition library, and historical data is called for index calculation; the method comprises the following steps of: according to stability, reusability and economic indexes such as a variable coefficient of a prediction error, a model calling frequency, economic benefit influence and the like, normalizing the indexes by adopting a range method, and calculating the weight of each index by utilizing an entropy weight method; constructing a weighted standardization matrix, defining a dynamic ideal solution, calculating a close degree index, and outputting a comprehensive evaluation result; according to the scoring result, an optimization instruction is automatically pushed, and model parameter adjustment or data calibration is carried out, a set of dynamic and multi-dimensional thermal power plant supervision model evaluation system is constructed, the model quality can be objectively and comprehensively evaluated, and a closed-loop optimization mechanism is provided.

The present invention relates to the data updating technology, disclosing a method for constructing a quality evaluation data updating model based on a convolutional neural network including: collecting historical data of product quality evaluation from an original quality evaluation model, determining the update frequency of quality evaluation data of original model; obtaining the latest quality evaluation data for data update of model based on the historical data of product quality evaluation; establishing a first data sample set and a second data sample set to update the model; and conducting model performance testing on the original model with such updated data to determine the effectiveness evaluation results of the data updates. The present invention updates the data samples of the quality evaluation model by determining the data update frequency of the model, and improves the performance and accuracy of the model by evaluating the effectiveness of the data updates.

The embodiment of the invention provides a quality evaluation method and device, a storage medium and a computer program product. The method comprises the steps of obtaining first data; wherein the first data comprises point cloud data corresponding to one or more Three Dimensional (3D) models, and the point cloud data comprises point cloud data corresponding to one or more Three Dimensional (3D) models; generating a quality evaluation result corresponding to the 3D model based on the point cloud data and the first model; wherein the first model comprises a first module, and the first module is at least used for converting the point cloud data into a first representation form and a second representation form; wherein the first representation form comprises a voxel grid form, and the second representation form comprises a two-dimensional (Two Dimensional, 2D) image form, that is, according to the embodiment of the invention, the quality of the 3D model can be evaluated based on the voxel grid data and the 2D image data at the same time, so that the advantages of the multi-modal data can be fully utilized, and the accuracy of the quality evaluation of the 3D model is improved.

The invention provides a quality control predictive analysis method and platform, and belongs to the technical field of predictive evaluation, and the method specifically comprises the steps: carrying out the evaluation processing of the quality score of an updated credit model in a predictive analysis period, so as to update the evaluation analysis results of the quality scores of the updated credit model in different predictive analysis periods, according to the control method for determining the model quality of the credit loan model, the matching condition of the credit loan model in a prediction analysis period between different update processing times is combined with the control method for the model quality of the credit loan model with different update processing times. And a prediction analysis method for quality control after the next update processing of the credit model is determined, so that the model quality of the credit model after the update processing is improved.

The invention discloses an automatic driving-oriented personalized federal learning and communication method, which is characterized in that local model and global model aggregation is carried out based on a parameter similarity adaptive algorithm, weighted aggregation is carried out by measuring the weight similarity of personalized local models uploaded by clients, and the problem of data isomerism is solved. And secondly, compressing the local model to obtain a teacher model and a student model, and cooperatively optimizing the teacher model and the student model through mutual distillation mechanism bidirectional knowledge migration to solve the problem of overlarge model quality and precision and communication overhead. And finally, model parameters are compressed based on singular value decomposition, so that a communication mode in personalized federated learning is constructed, and the method can be suitable for various complex environments.

The invention provides a CAE modeling and grid dividing method and system based on natural language driving, and the method comprises the steps: constructing a triple language model, defining an action-object-adversity triple structure, receiving a natural language instruction, carrying out the standardization processing of the instruction, recognizing and extracting the action, object and adversity components in the instruction, and carrying out the recognition and extraction of the action, object and adversity components in the instruction. The missing necessary parameters are prompted and complemented; based on a predefined analysis rule, extracting a triple conforming to the triple structure from the processed instruction; mapping the triad passing the verification into a corresponding CAE operation instruction through a mapping rule base, and generating an instruction sequence according to a logic sequence of a CAE modeling process; and automatically executing the instruction sequence, completing CAE modeling and grid division operations, and finally outputting a CAE model file. According to the method, the user operation complexity and the professional threshold are remarkably reduced, the modeling efficiency and the model quality are improved, an efficient and intelligent auxiliary tool is provided for engineering design, and the wide application value is achieved.

A gas temperature field reconstruction method using thermocouple measurement correction is disclosed. First, an Inventor 3D model and an ANSYS temperature field model are established for the gas temperature field. Then, a mapping relationship is established between the control parameters of the gas temperature field and the boundary conditions of the ANSYS temperature field model. Based on the principle of maximizing the influence of fuzzy boundary conditions on the temperature field reconstruction results, experiments are designed for precise boundary conditions and experimental data are obtained. Next, given the boundary conditions of the ANSYS temperature field model, the model is run to obtain the temperature field reconstruction results. A model quality assessment algorithm is run to calculate the fitting degree between the measured curve and the simulation curve, obtaining the algorithm fitting parameters. The algorithm fitting parameters are selected to form the loss function of the gradient descent algorithm. The learning rate, termination condition, and initial iteration parameters are determined. The gradient descent algorithm is run until convergence. Finally, the final iteration parameters are taken as the correction result of the fuzzy boundary conditions of the ANSYS temperature field model. This invention improves the spatiotemporal resolution of the temperature field reconstruction results.

The invention relates to the technical field of live-action three-dimensional modeling, in particular to a live-action three-dimensional model quality evaluation method, which comprises the following steps of: firstly, constructing a multi-dimensional evaluation index set covering dimensions such as geometric accuracy and texture quality; an initial weight is obtained through coupling of an analytic hierarchy process and an entropy weight method, and a dynamic weight is formed by combining scene demand coefficient adjustment; then, a subjective-objective linkage relation is established through grey correlation analysis, and index scores and weights are corrected through feedback coefficients; then calculating the total quality score of a single time node; the evaluation effectiveness in the dynamic scene is verified through a time sequence consistency index; and finally outputting the comprehensive quality grade. According to the method, a multi-dimensional index set is constructed to cover full dimensions, dynamic weights are calculated in combination with scene demand coefficients, and the problems of single dimension and poor scene adaptation are solved; building subjective-objective linkage by means of grey relational degree, correcting scores and weights, and solving disjunction; and calculating a time sequence consistency index to verify dynamic coherence, solving static limitation, and adapting to full-life-cycle management and control of the model.

The invention relates to a model training method and device, a target detection method and device and electronic equipment, and belongs to the technical field of neural network models.The model training method comprises the steps that a non-target domain training data set and target domain data are obtained, the non-target domain training data set is adopted to train migratable features of a to-be-trained model, and the to-be-trained model is obtained; obtaining a pre-training model; processing the target domain data by adopting the pre-training model to obtain a pseudo tag of each target domain data, and generating a target domain training data set based on each target domain data and the pseudo tag; and carrying out multi-batch training on the pre-training model by adopting the target domain training data set to obtain a trained model. According to the method, a large amount of high-quality target domain training data can be obtained, the model is trained, the quality of the model is improved, the model is trained through the non-target domain data with the label, and tasks such as identification and classification of the target domain data without the label can be achieved.

The invention relates to the technical field of building construction, in particular to an AI-based BIM (Building Information Modeling) and design budget integration method, which comprises the following steps of: firstly, constructing a high-precision enhanced BIM by fusing laser scanning point cloud, texture data and IoT (Internet of Things) environment data with a BIM original model; then, the trained multi-task deep learning model is utilized to automatically identify defects such as equipment missing, redundancy and geometric distortion in the enhanced BIM model, and a mapping relation between components and cost items is synchronously constructed; and then, calling a predefined strategy or a generative AI to repair the model according to the defect information, and updating a bill of quantity and a budget report in real time in a linkage manner. And finally, through difference degree evaluation and reinforcement learning driven closed-loop optimization, a compliant repair scheme with the lowest cost is automatically selected. According to the method, full-process automation and optimization from defect identification and intelligent repair to cost synchronization are realized, and the model quality and the cost control precision in the design stage are remarkably improved.

A set of building-oriented ship three-dimensional model modeling quality inspection set system is systematically constructed based on a 3DE platform, a corresponding model quality inspection method is provided for each inspection item, a set of three-dimensional model modeling quality inspection system is formed, information quality inspection for a ship product three-dimensional model is realized, and the inspection efficiency of the ship product three-dimensional model is improved. A designer is helped to improve the quality and efficiency of product design, discover model quality defects, reduce design repetition and promote ship product quality improvement.

The application belongs to the technical field of UML model fragment reuse, and relates to a UML model fragment reuse method, which filters out non-packable elements and external reference elements according to UML elements selected or input in other manners, recursively queries reference elements associated with valid packable elements, and completely collects all elements required to express a UML model fragment, thereby solving the core problems that the prior art cannot completely process associated metadata and reference relationships and cannot reuse UML model fragments across scenes. In addition, the UML model fragment template created is stored in an EMF resource, and relies on the standardized storage characteristics of EMF, so that the template can be conveniently managed and called across projects, and is compatible with other modeling tools. Finally, the UML model fragment is conveniently reused, and modeling efficiency, model quality and maintainability are significantly improved.

The present invention provides a quality control prediction and analysis method and platform, which belongs to the field of prediction and evaluation technology, and specifically includes: evaluating the quality score of the credit model after the update processing in the prediction and analysis period, determining the model quality control method of the credit model based on the evaluation and analysis results of the quality score of the updated credit model in different prediction and analysis periods, determining the prediction and analysis method for quality control of the credit model after the next update processing based on the matching of the prediction and analysis periods between different update processing times of the credit model, and combining the model quality control method of the credit model at different update processing times, thereby improving the model quality of the credit model after the update processing.

The invention relates to the technical field of material simulation, and discloses a closed-cell foamed aluminum acoustic performance simulation method based on a Voronoi model, which comprises the following steps: model construction: constructing a three-dimensional solid model of closed-cell foamed aluminum with adjustable porosity, aperture and wall thickness parameters based on the Voronoi principle; performing model optimization: performing homogenization processing on a Voronoi polyhedral structure in the three-dimensional entity model by adopting a random point homogenization algorithm so as to improve the quality of the model; and acoustic simulation: importing the optimized three-dimensional solid model into finite element simulation software, building a simulation model, and calculating the sound absorption performance or sound insulation performance of the simulation model. According to the method, the technical problems that an existing simulation model is distorted and the grid quality is poor are solved, the accuracy and reliability of a simulation result are remarkably improved by constructing the solid model which is accurate in geometry and uniform in structure, and a set of complete and verifiable technical method is provided for acoustic performance prediction and structural design of the closed-cell foamed aluminum.

The embodiment of the application discloses a model quality evaluation method, device, electronic equipment, product and storage medium, the method comprises the following steps: obtaining the source file of the building information model and the engineering construction specification information corresponding to the building information model; at least one specification entity and the association relationship between the specification entities are extracted from the engineering construction specification information to construct a target knowledge graph; information associated with at least one building component in the building information model is extracted from the source file to obtain building component association information; the building component association information is converted into knowledge structured data matching the data format of the target knowledge graph; the quality of the building information model is evaluated according to the knowledge structured data and the target knowledge graph, and an evaluation result is obtained. The scheme in the application can improve the efficiency and accuracy of the building information model quality evaluation result.

A method for evaluating a selection function for an active learning method of training a data-based model includes (i) providing training data sets and validation data sets which each associate an input data point with a label, (ii) performing a training of the data-based model using an active learning method on the basis of the selection function based on the training data sets, (iii) generating multiple evaluation quantities of test data sets by resampling from the validation data sets, (iv) determining a model quality and a level of uncertainty for the model quality on the basis of a statistical evaluation of the model performance of the data-based model based on the generated test data sets, and (v) maintaining or discarding the selection function based on the model quality and level of uncertainty.

The application provides a limit test method, system, equipment and medium of an artificial intelligence model, and the method specifically comprises the following steps: constructing a plurality of scene templates for different types of games, generating a test case set by combining and transforming each scene template according to a boundary condition input by a user; based on the test case set, determining an optimal test parameter combination satisfying the artificial intelligence model by using a heuristic search algorithm according to a preset optimization objective function; automatically testing the artificial intelligence model according to the optimal test parameter combination, generating a model test result; determining the performance of the artificial intelligence model under the boundary condition by analyzing the model test result, generating a model test report, and optimizing and retesting the artificial intelligence model through the model test report. The application realizes full-process automation and intelligentization of game AI model testing, and significantly improves the test efficiency and model quality.

This invention relates to the field of large model algorithm technology, and in particular to a dynamic compression optimization method and system for large models based on sparse pruning. The method captures raw weight fluctuation data caused by resource fluctuations during inference, and obtains sparse weight baseline data through sparsification. It analyzes the computational complexity of model inference latency data to separate the inference latency caused by model size. Based on the latency and sparse baseline data, the model compression ratio is dynamically controlled within a preset performance range, and inference accuracy distribution data under different compression parameters are collected. The model performance status under each parameter is evaluated using neural network simulation methods, generating performance status simulation results. Combined with real-time resource fluctuation data acquired during compression, model quality optimization compensation parameters are determined based on the simulation results. Finally, the compression strategy is adaptively adjusted through the compensation parameters to achieve coordinated optimization of model computational complexity, inference accuracy, and latency when hardware resources change dynamically.

The invention provides an incentive and secure federal learning method and system based on a block chain, and the method comprises the steps: a plurality of participants obtain training tasks from the block chain, submit corresponding bidding prices to a smart contract, the smart contract determines a winner set in current training according to the bidding prices, and randomly divides the winner set into a training set and a verification set, obtaining a corresponding trainer and a verifier; the method comprises the following steps: downloading a currently trained local model from a block chain based on winner, adding artificial Gaussian noise, calculating a noise gradient of a target function of the local model by each winner, transmitting the noise gradient to each verifier for model verification, obtaining a list of the selected local model, and performing global aggregation; all verifiers are divided into small fragments, after a verification task of a local model of the trainer is received, each verifier in the small fragments evaluates the quality of the model, and the verifiers exchange votes mutually. The fragmentation consensus algorithm based on circulation can ensure the same security as a non-fragmentation consensus protocol.

The invention provides a digital twinborn modeling method and system for a distribution network tower. The method comprises the following steps: carrying out classified statistical analysis on geometric structure characteristic data of a target tower to determine a parameter distribution proportion of each component; modeling samples are selected from the tower point cloud database to construct a three-dimensional modeling sample set, and it is ensured that component parameter distribution in the sample set is consistent with the statistical proportion; and generating a model construction instruction based on the sample set, driving a digital twin modeling engine to perform tower reconstruction, and collecting performance data in a modeling process to generate a quality evaluation index. According to the method, accurate and efficient reconstruction of the tower model is realized through parameterized sample selection and standardized instruction generation, and a quantitative evaluation basis is provided for modeling quality. According to the method, the construction precision and efficiency of the digital twinborn model of the distribution network tower can be improved, and modeling evaluation and optimization are supported.

The invention relates to the technical field of airborne system design, and discloses an airborne system model modeling rule making method and device, equipment and a medium, which can deeply fuse airborne system design multi-source heterogeneous knowledge, automatically generate executable modeling rules, improve airborne system design efficiency and model quality, and improve airborne system design efficiency and model quality. The problem that modeling rules are inconsistent due to the fact that traditional modeling rules depend on artificial experience is solved, and standardization, consistency and quality control efficiency of airborne system design are improved.

A system and method are provided for fine-tuning large language models (LLMs). The method provides high system performance during fine-tuning and attains state-of-the-art model quality on downstream applications. The method identifies the most important LLM weights via second-order information in a pre-processing step, and significantly reduces the computation and storage costs of the pre-processing step via i) a hierarchical approximation of second-order information, and ii) an online projection and rediagonalization algorithm. The method may train only the sparse important weights and embeds these sparse weights into the pre-trained LLM during fine-tuning to provide high system performance.

The invention discloses a parameterization and modularization combined ancient building modeling method and system, and the method comprises the steps: constructing a modular component library which comprises a plurality of parameterization control components as minimum building units, and an ancient building universal module formed by a plurality of parameterization control components; constructing a modular material mapping library, and obtaining historic building material texture information to construct a mapping unit; obtaining target characteristic parameters of the target historic building; calling a target building unit corresponding to the target historic building from the modular component library; and according to the target characteristic parameters, performing parameterized control modeling on the called target building units by using a parameterization algorithm, optimizing a connection structure between the target building units by using an intelligent adaptation algorithm, and calling mapping units from the modular material mapping library to perform mapping processing on the preliminary model to obtain a target historic building three-dimensional model. According to the method, accurate adjustment of the parts is realized through parameterization control, and the modeling efficiency and the model quality are greatly improved.

The invention discloses a method and equipment for improving engineering problem modeling quality and a storage medium, and belongs to the field of engineering optimization problem modeling by a large language model. The method comprises the following steps: step 1, organizing a series of engineering optimization problems according to hierarchical classification and complexity, and constructing a tree structure knowledge base; step 2, receiving natural language description of a target engineering problem to be modeled, performing recursive search on the tree structure knowledge base in the step 1, and identifying modeling sub-problems corresponding to nodes related to the target engineering problem until the most matched and most specific modeling sub-problem is found; 3, retrieving an advanced modeling thought corresponding to the modeling sub-problem in the step 2, and combining the advanced modeling thought with the description of the target engineering problem to generate a global modeling thought; and 4, automatically generating an optimization model and solver codes of the target engineering problem by using the large language model according to the global modeling thought in the step 3. According to the method, the automatic modeling quality of solving a complex engineering optimization problem by using a large language model can be improved.

Examples described herein relate to systems and methods for automatic evaluation of graphical element recommendations, such as sticker recommendations. According to some examples, a system accesses a set of text queries and provides each text query as input to a graphical element recommendation machine learning model. The graphical element recommendation machine learning model is trained to generate, based on a given text query, one or more graphical element recommendations for use in a message in a context of a messaging interface of an interaction application. The system obtains, from the graphical element recommendation machine learning model, at least one graphical element recommendation for each text query. The system generates a model quality score for the graphical element recommendation machine learning model by applying a model quality metric to the graphical element recommendations. Output indicative of the model quality score may be presented at a user device.

The invention discloses a multi-view-based three-dimensional reconstruction model quality evaluation method and system and a medium, and belongs to the field of multi-view three-dimensional reconstruction model quality evaluation. Comprising the following steps that a model generated through multi-view three-dimensional reconstruction is obtained, and the surface of the model is a Lambert body surface; the rendering equation of the Lambert body surface is converted into a point multiplication form of an incident light spherical harmonic coefficient and a surface transmission characteristic spherical harmonic coefficient, and the surface transmission characteristic is a product of a binary visible function and a geometric cosine term; the real irradiance of the surface of the reconstruction model is reversely deduced through the pixel intensity of the multi-view image; based on the incident light spherical harmonic coefficient, calculating the reconstruction irradiance of the reconstruction model by using a spherical harmonic coefficient linear combination equation; if the difference value between the reconstruction irradiance gradient and the real irradiance gradient between any two points in the model is greater than a preset threshold value, determining that a geometric error exists in a three-dimensional reconstruction result between the two points; according to the method, which parts of the reconstruction model have errors can be accurately judged.