296 results about "Model validation" patented technology

The invention discloses a coal rock fracture intelligent extraction method based on improved U-Net. The method comprises the following steps: S1, constructing a coal rock fracture CT image data set; s2, constructing an improved U-Net segmentation model, specifically comprising the following steps: S2.1, taking VGG16 as a backbone network, and introducing a depth separable convolution module; s2.2, a PPA attention module is added after each layer of depth separable convolution of the decoder, the PPA attention module is introduced after each up-sampling stage of the decoder, and the output of the PPA attention module is subjected to batch normalization and Dropout layer processing; s2.3, defining a composite loss function; s3, training and optimizing a segmentation model, wherein the specific steps comprise: S3.1, setting hyper-parameters; and S3.2, training the model by using the training set, adjusting hyper-parameters by using the verification set, and evaluating the performance by using the test set, wherein the evaluation indexes comprise MIoU, MAcc and FWIoU. According to the method, the problems of difficult identification of small fractures, large model calculation amount, poor multi-scale information fusion and class imbalance in the coal rock fracture image can be solved, and the robustness, segmentation precision and practicability of the model are improved.

The invention relates to a method for constructing a high-resolution atmosphere carbon dioxide concentration data set based on XGBoost-BO, and belongs to the technical field of environment monitoring and artificial intelligence modeling. The method comprises the following steps: preprocessing OCO-2 satellite data and multi-source auxiliary data, and fusing the preprocessed OCO-2 satellite data and multi-source auxiliary data to obtain a new data set; a Bayesian optimization method is adopted to search for an optimal hyper-parameter, a target function is optimized through second-order Taylor expansion, a regular term is introduced to control the complexity of the model, and ten-fold cross validation is used to evaluate the performance of the model; quantizing the contribution degree of each feature to model prediction through a tree SHAP method, and analyzing global feature importance ranking and feature contribution distribution of individual samples; and performing model verification by using the test set and the site actual measurement data. According to the method, the problems that an existing model-based reconstruction method is insufficient in interpretation and prone to falling into local optimum are solved, and the temporal-spatial resolution of CO2 concentration monitoring can be improved.

The invention relates to the technical field of ship energy consumption prediction, and discloses a ship energy consumption interval prediction method and system based on a Gaussian quantile regression model.The method comprises the steps that firstly, a ship navigation historical data set is preprocessed, a model input feature set is screened, a Gaussian process quantile regression model with a radial basis function as a kernel is constructed, and hyper-parameters are optimized; and after the prediction performance of the subset evaluation point is verified through the model, an upper quantile prediction model and a lower quantile prediction model are respectively established according to a target confidence level, and finally a prediction interval is synchronously output and an evaluation report is generated. According to the method, through combination of Gaussian process processing nonlinear relation and quantile regression estimation condition distribution, high-precision point prediction is provided, meanwhile, prediction uncertainty can be quantized, an energy consumption prediction interval corresponding to a target confidence level is output, and more comprehensive and reliable information support is provided for ship energy efficiency management and operation decision making.

The invention discloses a verifiable privacy protection federated learning method based on sensitive samples, and relates to the field of fault diagnosis. According to the method, the Poisson sampling process is introduced, the sampling probability is generated based on the privacy budget, it is ensured that all recorded privacy budgets are synchronously exhausted, data disastrous forgetting is effectively prevented, and the model effectiveness is improved. In the model verification module, a sensitive sample set is generated by means of a gradient maximization algorithm, a model integrity attack is detected, a user side only needs to submit a small number of sensitive samples for prediction through an inference service API provided by a private cloud client side, and if a returned result and a real result have significant deviation, it can be judged that the model is possibly tampered. According to the method, diversified privacy requirements of users can be met, black box verification on the integrity of the model is realized, and the precision of the model is improved.

The invention discloses an anti-money laundering risk rule generation method and system based on a multi-modal large language model, and the method comprises the following steps: obtaining an initial training set composed of transaction data, adding a money laundering label based on a rule model, extracting transaction features through a multi-modal large language feature extraction model, and carrying out the expansion, forming a modeling training set by the initial training set and the expanded data set; fitting the multi-modal large language learning prediction model through the modeling training set to obtain a money laundering risk value of the transaction data, and calculating an actual contribution value of the structured features; and inputting the verification set into the multi-modal large language learning prediction model, verifying to obtain a money laundering label re-determination result, updating the money laundering label re-determination result to the initial training set, and then generating an updated rule model according to the verification result and the contribution value of the structured feature to the money laundering risk value. According to the invention, multi-mode transaction data can be integrated, a risk prediction model of money laundering transaction is formed, real-time adjustment and updating are carried out, and money laundering transaction behaviors can be accurately identified.

The invention discloses a compound structure identification database construction method, system and device and a medium, and the method comprises the steps: inputting a model training data set into a plurality of retention time prediction models for training, and obtaining a plurality of trained retention time prediction models; verifying the plurality of trained retention time prediction models by adopting a model verification data set, and selecting an optimal retention time prediction model; inputting the migration training data set and the migration verification data set into the optimal retention time prediction model for migration learning to obtain a target retention time prediction model; inputting to-be-predicted compound structure data into the target retention time prediction model for prediction to obtain predicted retention time; and constructing a compound structure identification database according to the predicted retention time. The data in the compound structure identification database can be perfected, and the accuracy of the compound structure identification database is improved, so that the database performance is improved.

Disclosed herein are systems and methods for cross-domain training of sensing-system-model instances. In an embodiment, a system receives, via a first application programming interface (API), an input-dataset selection identifying an input dataset, which includes a plurality of dataframes that are in a first dataframe format and that have annotations corresponding to one or more sensing tasks performed with respect to the dataframes. The system executes a plurality of dataframe-transformation functions to convert the plurality of dataframes of the input dataset into a predetermined dataframe format. The system trains an instance of a first machine-learning model using the converted dataframes of the input dataset to perform at least a subset of the one or more sensing tasks. The system outputs, via the first API, one or more model-validation metrics pertaining to the training of the instance of the first machine-learning model.

The invention discloses a register model and register transfer level generation and test method, device, equipment and medium, and relates to the technical field of integrated circuit design verification, and the method comprises the following steps: analyzing a comma separation value table file by using a register generation platform to generate a register model, a register transfer level and a register test vector; the comma separation value table file is a file used for describing register information; generating a general verification methodology environment by using a test environment generation platform, and integrating the general verification methodology environment based on the register model, the register transfer level and the register test vector to obtain a corresponding integrated test environment; and compiling the integrated test environment, if the compiling is successful, running the register test vector in the integrated test environment which is successfully compiled, and outputting a corresponding test result. Therefore, the automation degree of register model verification can be improved, and the verification efficiency of front-end verification is improved.

The systems and methods disclosed herein relate to a model validation platform that enables dynamic validation of a user's prompt for a large language model (LLM) in order to evaluate the validity of the prompt and the suitability of a large language model for processing the prompt. For example, the platform enables an estimation of the resource allocation associated with processing the prompt with a given LLM, as well as a modification of the prompt, prior to the processing the prompt with the selected LLM. The platform can further validate the output prior to transmitting the output to a server system for display to the user. By doing so, the platform enables dynamic evaluation of a request to execute an LLM, as well as evaluation of resulting outputs, for accuracy and efficiency improvements in data processing or software development pipelines.

The invention discloses a target detection online learning dynamic sample selection method and system, computer equipment and a storage medium. The method comprises the steps that classification uncertainty and positioning uncertainty of samples to be screened are obtained through Monte Carlo Dropout sampling; constructing multi-dimensional feature vectors including classification uncertainty, positioning uncertainty, knowledge gap matching degree and the like; constructing a dynamic weight learning network based on an attention mechanism, and calculating the weight of each feature in combination with a model verification set performance index; and performing multi-dimensional value scoring on the samples according to the feature weights, and screening out an optimal sample subset in combination with calculation power limitation so as to complete online updating of the model. According to the method, the sample value is comprehensively evaluated through multi-dimensional feature fusion, different scene requirements are adapted by utilizing dynamic weights, and resource consumption and updating effects are balanced in combination with computing power perception sampling, so that the adaptability and detection precision of the model in a dynamic scene are effectively improved, and meanwhile, the dependence on manual annotation is reduced.

The invention discloses a fully-weathered argillaceous siltstone elastic-plastic damage constitutive model establishing system, and relates to the technical field of elastic-plastic damage constitutive models.The system comprises a parameter collecting and labeling module used for obtaining a data set and conducting classification labeling and test condition information; the parameter processing and matching module is used for preprocessing the parameters to extract a feature parameter set and performing information association to supplement boundary condition data; the damage elastic-plastic feature module is used for presetting a database and a weight distribution rule and performing weighted fusion to form a damage elastic-plastic coupling feature parameter set; the constitutive model construction module is used for constructing a constitutive model, performing iterative training on the constitutive model and calculating damage parameters and a prediction result; and the model verification report module is used for verifying the damage parameters and the prediction result and judging the effectiveness of the constitutive model according to the verification result. According to the method, the multi-dimensional core characteristic parameter set is constructed by obtaining mechanical property test parameters and the like, and the limitation of single test data is broken through.

The invention relates to the technical field of rail transit construction, and discloses an analogue simulation method for a shield mud film gas check test, which comprises the following steps: S1, establishing an analogue simulation system for the shield mud film gas check test; s2, preparing slurry according to requirements; s3, measuring the porosity, density, residual moisture content, water conductivity coefficient, elastic modulus, Poisson's ratio and water holding capacity parameters of the slurry and the stratum soil body; s4, all the material parameters obtained in the S3 are substituted into an analogue simulation system of the shield mud film air closing test, and then a small amount of mud film air closing tests are used for carrying out model verification; and S5, carrying out simulation calculation analysis. According to the method, an indoor gas closing test simulation model is established on the basis of a Biot pore elasticity theory under a fluid-solid coupling condition, different water holding capacity characteristics of a mud film and a stratum are considered, and model verification is carried out through a gas migration rule and leakage flow evolution data obtained through an indoor test; and the rapid evaluation of the closeness characteristic of the mud film under complex geological conditions and shield parameter working conditions is realized.

The invention discloses a cross-scene wastewater treatment control system based on knowledge distillation and rule migration and an implementation method, and relates to intelligent control of industrial wastewater treatment. The method comprises the following steps: collecting operation parameters, control rules and wastewater treatment effects of a plurality of wastewater treatment systems, and completing data preprocessing; then performing feature extraction on the wastewater treatment process and the control rule by utilizing a knowledge distillation technology, and constructing a standardized multi-system data set; constructing a knowledge distillation base model for data fusion of different processing systems, and performing model verification; constructing an interpretable general rule set, namely calculating a feature matching degree of the general rule set in a new scene, and performing rule migration correction by utilizing a specific small sample; and substituting the generated target system rule into a triple discriminator for verification, and determining a final deployable wastewater treatment system control rule. According to the method, a machine learning model driven by multi-source data is converted into a general interpretable rule base through a regular distillation algorithm, and intelligent control over lossless industrial deployment is achieved.

The invention discloses a multi-Agent cooperation AADL model AGREE verification statement automatic generation and verification method. The method comprises the following steps: S1, constructing a domain knowledge base; s2, the Agent is analyzed by the AADL model to extract key information; s3, the demand analysis Agent splits atomic propositions from the natural language demand; s4, the AGREE generator Agent generates an AGREE attachment code; s5, the model fusion Agent automatically embeds the AGREE specification into the target AADL component; s6, the AADL verification Agent performs iterative verification and automatic repair, and a multi-dimensional verification system is constructed; and S7, executing AGREE-based AADL architecture formalized verification, and performing logic attribute verification on the grammar compliance model. According to the invention, innovative fusion of a formalized verification method and an intelligent technology is realized, and the core problem that formalized verification of the AADL architecture is difficult to implement is effectively solved.

The invention discloses a model robustness automatic evaluation method based on large model driving and related equipment. The method comprises the following steps: inputting a task template of a tested model into a dynamic Prompt engine module to obtain a sample generation instruction; generating an adversarial sample according to the sample generation instruction through a sample generation and verification module, and performing model evaluation on the adversarial sample; if the confrontation sample passes the model verification, extracting a high-dimensional multi-modal feedback vector of the tested model through a multi-modal feedback analysis module; a reusable semantic instruction is generated through a strategy control module according to the high-dimensional multi-modal feedback vector so as to optimize a generation strategy; and if the confrontation sample does not pass the model verification, performing effect evaluation on the confrontation sample through an evaluation and interpretation generation module and generating an interpretability evaluation report. According to the method, progressive generation of the adversarial sample and exponential-level improvement of the model robustness evaluation efficiency can be realized, a clear model improvement direction is provided for developers, and the method can be widely applied to the technical field of computers.

The invention relates to the technical field of electric power simulation, and discloses a hardware-in-the-loop simulation-oriented new energy equipment model verification method and system, and the method comprises the steps: constructing a new energy equipment basic model based on a physical equation, carrying out the compensation of a dynamic working condition error through combining with an LSTM neural network, and generating a hybrid parameterized model; fitting voltage and current data through a particle swarm optimization algorithm, and dynamically adjusting impedance parameters to complete model calibration; a time domain step response test, frequency domain characteristic analysis and a voltage sag fault experiment are executed on a hardware-in-loop platform, and the performance of the model is verified in a multi-dimensional mode; and generating a standardized verification report containing a steady-state error and a simulation step length index based on the deviation between the dynamic response curve and the measured data. According to the method, a mechanism model and a data driving technology are fused, the problems that dynamic characteristic adaptation is insufficient and parameter calibration depends on experience in traditional verification are solved, and the simulation precision and verification efficiency of the model under complex working conditions are improved.

Conventional estimation model generating apparatuses are required to be able to estimate information on a flow rate of a target river. An estimation model generating apparatus 101 includes: a basin data acquiring unit 142 that acquires basin data of a basin of a river of interest from an image containing the basin; a candidate model acquiring unit 143 that acquires one or more candidate models for estimating information on a flow rate of the river based on the basin data; a verifying unit 145 that verifies validity of at least one candidate model out of the one or more candidate models using verification data; an estimation model acquiring unit 147 that acquires one candidate model out of the one or more candidate models as an estimation model based on a verification result; and a model accumulating unit 149 that accumulates the acquired estimation mode. Accordingly, it is possible to estimate information on a flow rate of a target river.

The systems and methods disclosed herein enable evaluation of machine learning model outputs within a virtual environment. The disclosed model validation platform enables testing of code generated for detection of malicious or anomalous outputs. For example, the model validation platform can construct a virtual machine isolated from the system and test model-generated code for validation of LLM-generated outputs. In some implementations, the model validation platform determines parameters of the virtual machine and / or associated validation test based on an evaluation of the machine learning model's output and / or the associated underlying prompt. For example, the parameters of the validation test depend on an evaluation of the user or the provided input (e.g., depending on the presence of sensitive data within the prompt). By doing so, the system enables dynamic evaluation of machine learning model outputs to improve the security and robustness of associated generated code.

The invention relates to the field of power system modeling and parameter identification, and provides a synchronous generator parameter identification method based on an improved tuna swarm optimization algorithm. Dividing the full parameter set of the to-be-identified synchronous generator into a d-axis parameter subset, a q-axis parameter subset and a mechanical parameter subset according to the internal physical association of the to-be-identified synchronous generator; performing multiple improvements on the tuna swarm algorithm to obtain an improved tuna swarm optimization algorithm; integrating the to-be-identified parameters in the d-axis parameter subset and the q-axis parameter subset into a high-dimensional parameter vector, constructing a target function based on the comprehensive test data, and performing multi-parameter synchronous global identification based on the target function by adopting an improved tuna swarm optimization algorithm to obtain an identification result; and carrying out model verification and validity analysis on the identification result. According to the method, the problem of parameter correlation distortion of traditional step-by-step identification is solved, the global optimization ability and convergence efficiency of the algorithm are improved, and the method can adapt to the accurate identification requirement of the high-dimensional strong coupling parameters of the synchronous generator.

The present invention relates to a multi-objective fuel cell cooling channel optimization design method, comprising: utilizing control equation coupling to construct a two-dimensional, two-phase, non-isothermal fuel cell mathematical model, which, after model validation, serves as a data-driven source; extracting cooling channel structural optimization parameters, optimization objectives, optimization parameter variation ranges, and constraints; substituting the varied optimization parameters into the mathematical model based on the optimization parameter variation ranges and constraints to output an original data set; applying a machine learning algorithm based on the original data set to construct a data-driven, multi-objective proxy model; and optimizing and solving the multi-objective proxy model using a genetic algorithm to obtain optimal cooling channel structural parameters. Compared with existing technologies, the present invention can comprehensively optimize cooling channel structural parameters from multiple angles, quickly, and accurately, providing guidance for practical fuel cell structural design.

The invention provides an inverse constraint inference method and device, equipment and a medium, and relates to the technical field of artificial intelligence. Comprising the steps of obtaining a diffusion generation model verifier and a reward model through pre-training on a data set containing expert and non-expert demonstration, and freezing model parameters of the diffusion generation model verifier and the reward model; obtaining a first track generated by a diffusion generation model verifier under the guidance of the constraint model and the reward model, and updating the constraint model according to a comparison learning result of the first track and an expert data track in the data set to obtain a first constraint model; and continuously obtaining a second track generated by the diffusion generation model verifier under the guidance of the first constraint model and the reward model, and updating the first constraint model according to a comparison learning result of the second track and the expert data track in the data set by adjusting a loss item coefficient of comparison learning. And obtaining a second constraint model with different loss item coefficient conditions. According to the invention, the flexibility and efficiency of inverse constraint inference can be improved.

Systems and methods for federated model validation and data verification are disclosed. A method may include: (1) receiving, by a local computer program executed by client system, a federated machine learning model from a federated model server; (2) testing, by the local computer program and using a policy service, the federated machine learning model for vulnerabilities to attacks; (3) accepting, by the local computer program, the federated machine learning model in response to the federated machine learning model passing the testing; (4) training, by the local computer program, the federated machine learning model using input data comprising local data and outputting training parameters; (5) identifying, by the local computer program using the policy service, accidental leakage and / or contamination by comparing the training parameters to the input data; and (6) providing, by the local computer program, the training parameters to the federated model server.

The invention relates to the technical field of environmental risk management and control, in particular to a tire wear particle environment fate prediction model construction method based on multi-dimensional data. Comprising the following steps: data acquisition and preprocessing; a feature importance analysis model based on a random forest is constructed, key factor mining is performed on the preprocessed collected data, core factors influencing tire wear particle environment fate are screened by calculating Gini importance values of the key factors, and feature vectors and confidence coefficients of the core factors are output; classifying and integrating factors; carrying out model association construction; model training storage; and performing model verification feedback. According to the method, the continuity, timeliness and consistency of tire wear particle sample data and environment associated data are improved through the built tire wear particle environment fate prediction model and through cooperative collection, space-time registration and noise filtering preprocessing of the multi-source monitoring equipment, and a high-quality data basis is provided for subsequent model building.

The invention discloses a bulk cargo ship navigational speed prediction system based on feature engineering and an integrated tree model, and relates to the technical field of navigational speed prediction. The bulk cargo ship navigational speed prediction system based on the feature engineering and the integrated tree model comprises a feature optimization judgment and model training module, a model optimization judgment module, a model structure optimization judgment module and a navigational speed prediction output module. According to the method, whether bulk cargo ship characteristic optimization is carried out or not is judged through bulk cargo ship state characteristics, the optimized bulk cargo ship characteristics are input into a bulk cargo ship navigational speed prediction model for training, and then whether model parameter optimization is carried out or not is judged based on a model verification result; judging whether to perform model structure optimization or not based on a model verification result after model parameter optimization, and finally inputting real-time bulk cargo ship characteristics into the bulk cargo ship navigational speed prediction model after model parameter optimization to obtain the bulk cargo ship navigational speed, thereby improving the effectiveness of the bulk cargo ship navigational speed prediction. The problem of low effectiveness of bulk cargo ship navigational speed prediction in the prior art is solved.

The invention discloses a multi-source heterogeneous data model modeling method and system, and belongs to the field of data modeling. The method solves the problem of modeling mainly aiming at a single data source or isomorphic data in the prior art, realizes global real-time information acquisition by accessing multiple types of data sources, can dynamically adjust an acquisition strategy according to data source characteristics and business requirements, ensures high efficiency and adaptability of data acquisition, and is suitable for large-scale popularization and application. Through data extraction and conversion, corresponding extraction strategies can be formulated for different types of data, formats are effectively cleaned and unified, the data are fused, a high-quality data basis is provided for modeling, during model construction, a proper algorithm is screened according to data characteristics, a model reflecting data relations and characteristics is accurately constructed, and then through model verification and optimization, a high-quality data basis is provided for modeling. The performance and generalization ability of the model can be improved, so that the high-quality, stable and reliable model is constructed.

A method may include receiving, by a first user device from a network node or a second user device, 1) a positioning measurement report including at least one positioning measurement measured by a reference device, and 2) reference positioning-related information to be used for testing and / or validating a machine learning model; determining estimated positioning-related information as outputs of the machine learning model based on at least a portion of the positioning measurement report as inputs to the machine learning model; determining a performance indication of the machine learning model based on the reference positioning-related information and the estimated positioning-related information, wherein the performance indication indicates a performance or accuracy of the machine learning model; and performing, by the first user device, an action based on the performance indication.

The invention relates to the technical field of digital twinning, in particular to a digital twinning driven equipment information model construction system. Comprising a data acquisition and preprocessing unit; a digital twin modeling unit; the model verification and optimization unit adopts a virtual and real data real-time comparison mechanism and is combined with a self-adaptive optimization algorithm driven by deviation characteristics; and a model application and updating unit. A dynamic weight distribution algorithm is adopted in a model verification and optimization unit, the initial weight is adjusted based on the current load rate and the accumulative operation duration of equipment, and instantaneous fluctuation and trend deviation are distinguished through a two-dimensional fusion judgment algorithm in combination with the deviation duration and the trend slope. The problem of insufficient accuracy of virtual and real data comparison in the prior art is solved; according to different deviation properties, an adaptive genetic algorithm is adopted to generate an optimization strategy, and it is guaranteed that the digital twin equipment information model is consistent with the physical characteristics of an equipment entity for a long time.

Disclosed herein are systems and methods for cross-domain training of sensing-system-model instances. In an embodiment, a system receives, via a first application programming interface (API), an input-dataset selection identifying an input dataset, which includes a plurality of dataframes that are in a first dataframe format and that have annotations corresponding to one or more sensing tasks performed with respect to the dataframes. The system executes a plurality of dataframe-transformation functions to convert the plurality of dataframes of the input dataset into a predetermined dataframe format. The system trains an instance of a first machine-learning model using the converted dataframes of the input dataset to perform at least a subset of the one or more sensing tasks. The system outputs, via the first API, one or more model-validation metrics pertaining to the training of the instance of the first machine-learning model.

The invention relates to the technical field of battery detection, in particular to a battery health state estimation method, which comprises the following steps: S1, acquiring battery historical data including historical voltage, current, temperature and capacity data; s2, health factor extraction: analyzing the historical data of the battery to obtain SOH data, and extracting health factors closely related to SOH; s3, processing the health factors and the capacity data; s4, establishing an estimation model, adopting a dung beetle optimization algorithm to optimize hyper-parameters of the gating circulation unit, then introducing an attention mechanism, and establishing a TCN-DBO-GRU-Attention model as the estimation model; s5, model training; and S6, model verification and result evaluation. The method can accurately estimate the SOH of the battery, is suitable for batteries of different models, and has good accuracy and universality.

The invention relates to the technical field of computational fluid mechanics and biological microfluidic design, in particular to a model order reduction method for rapid optimization of a circulating tumor cell (CTC) sorting structure, which comprises the following steps of: constructing a three-dimensional full-order computational fluid mechanics model by collecting channel geometric structure parameters and fluid working condition parameters, and generating a training data set; organizing a flow field snapshot into a matrix form, extracting a dominant mode, constructing a low-dimensional modal space by taking the dominant mode as a base vector, establishing a low-dimensional ordinary differential model through Galerkin projection, and training a parameter-modal coefficient mapping relation by adopting a deep neural network to form a complete reduced-order model; and finally, constructing a multi-objective optimization problem based on the reduced-order model, carrying out optimization iteration by adopting an evolutionary algorithm, and returning an optimization result to the full-order model for verification, so that rapid optimization design of the CTC sorting structure is realized, and an effective solution is provided for intelligent design of a biomedical microfluidic device.