73 results about "Initial sample" patented technology

This application discloses a method, device, medium, and product for optimizing the structure of a shaped charge shroud, relating to the field of structural design. The method includes: constructing an objective function with structural parameters as design variables and maximizing performance index values ​​as the objective; determining multiple initial sample points using a Latin hypercube sampling method based on the range of structural parameter values; constructing an initial sample library; constructing a surrogate model based on the initial sample library; determining candidate sample points and their corresponding performance index values ​​using a Bayesian optimization loop based on the surrogate model and the range of structural parameter values; updating the initial sample library and the surrogate model; continuing until a termination condition is met; and using the sample point corresponding to the maximum performance index value as the target structural parameter; and optimizing the design of the shaped charge shroud based on the target structural parameter. This application can reduce the computational cost of determining the structural parameters of a shaped charge shroud and improve the efficiency and intelligence of shaped charge shroud structure optimization.

The application discloses a kind of sparse magnetic field data reconstruction method, system, equipment and storage medium, is related to magnetic field imaging and neural network image processing technical field, including by data acquisition department obtaining the initial sampling sparse magnetic field data of target detection area;Initial sampling sparse magnetic field data is input into sampling strategy department, generates sampling probability distribution graph, according to sampling probability distribution graph, obtains active sampling sparse magnetic field data;Active sampling sparse magnetic field data and initial sampling sparse magnetic field data are input into spatial superposition department, obtain final sampling sparse magnetic field data;Final sampling sparse magnetic field data is input into sparse data reconstruction module, and high-resolution reconstructed magnetic field data is output.The method described in the application is more good in improving sampling efficiency, optimizing reconstruction quality, reducing redundant data, enhancing reconstruction stability.

The application discloses a lightning impulse grounding grid frequency characteristic analysis method, system, device and medium, including: obtaining initial sample data; using the sample and a preset test frequency set to construct at least two different structure rational function interpolation models, and calculating the output residual thereof in the test frequency band; determining the target frequency point with the maximum modeling uncertainty according to the residual, calling the moment method to supplement high-precision response samples at the point; based on the updated sample set, constructing a diagonal or near-diagonal form rational interpolation model, and judging whether the prediction error of the model at the target frequency point is lower than a preset convergence threshold; if the convergence condition is not met, returning to the multiple model construction step, reiterating using the current sample set until the error meets the standard, and finally outputting the grounding grid frequency response characteristics covering the lightning impulse frequency band. The application realizes efficient analysis of the grounding grid frequency characteristics under lightning impulse, and effectively reduces the time used for analyzing the grounding grid frequency characteristics under lightning impulse.

Embodiments of the present application disclose a method, device, equipment, medium and program product for generating an adversarial sample. The method comprises: obtaining a preset code file; determining a non-conditional jump instruction in the preset code file; modifying the non-conditional jump instruction into a reconstructed jump instruction according to a preset instruction modification rule; performing control flow flattening processing on a basic block of the preset code file; based on a result of the control flow flattening processing, inserting a preset application programming interface (API) sequence into the preset code file according to a preset rule to obtain a target code file containing a malicious code feature; inserting noise into the target code file to obtain an initial sample; and performing iteration on the initial sample according to a preset evolution algorithm to obtain an adversarial sample. The embodiments of the present application can generate a large number of malicious code adversarial samples that can evade antivirus software and sandboxes.

The application relates to a parallel multi-objective optimization method and system for engineering product CAE simulation, and belongs to the technical field of simulation optimization. The method solves the problems that the simulation-driven design cycle is long and it is difficult to obtain a high-quality multi-objective trade-off optimization scheme. The method comprises the following steps: based on multiple objective functions and constraint conditions, initial sample data is obtained through two-stage collaborative optimization and is put into a sample library; a joint surrogate model is constructed and trained, and a hierarchical error compensation mechanism is initialized; multiple rounds of iterative optimization are performed until a preset optimization termination condition is met; each round of iterative optimization comprises the following steps: based on the current joint surrogate model and the hierarchical error compensation mechanism, multiple candidate design points are generated in parallel, and then the multiple candidate design points are distributed to multiple computing nodes for parallel simulation calculation to obtain new sample data which is put into the sample library; the joint surrogate model and the hierarchical error compensation mechanism are updated; and a multi-objective optimization design scheme set of a product to be optimized is obtained from the final sample library. The optimization efficiency is improved.

This invention discloses a small-sample Bayesian optimization sampling method and system based on cloud droplet data augmentation, belonging to the field of transportation engineering material design. Addressing the problems of limited initial samples and poor fit of the Bayesian optimization surrogate model in modified asphalt formulation design, this invention acquires initial small-sample data; constructs a clustering cloud model to obtain expectation, entropy estimates, and hyperentropy estimates; generates cloud droplet virtual samples using a forward cloud generator and merges them with the original samples to expand the dataset; trains a Gaussian process regression surrogate model based on the expanded data; constructs an expectation-improved acquisition function to optimize and solve candidate formulations; updates the data after physical testing verification; and iterates repeatedly until the termination condition is met to output the optimal formulation. This invention mines the distribution information of small-sample data through cloud droplet data augmentation, improves the accuracy and sampling efficiency of the surrogate model, significantly reduces the number of expensive physical tests, and lowers the cost of formulation design. It can be widely applied to the formulation optimization of modified asphalt and similar high-cost experimental materials.

Method and apparatus for training multimodal large model and method and apparatus for image question answering are disclosed, which relates to artificial intelligence technologies such as large models, deep learning, natural language processing, and computer vision. The method for training multimodal large model includes: obtaining an initial sample image, a sample object in the initial sample image, and a location information of the sample object; obtaining a target sample image including a sample visual marker based on the initial sample image and a target image region corresponding to the initial sample image; obtaining a sample question corresponding to the target sample image based on the sample visual marker, and obtaining a sample answer corresponding to the sample question; training an initial multimodal large model based on a target training sample constituted by the target sample image, the sample question and the sample answer to obtain a target multimodal large model. The method for image question answering includes: obtaining a target image including a target visual marker and a target question; inputting the target image and the target question into the target multimodal large model to obtain a target answer. The present disclosure enables the target multimodal large model to effectively understand the target visual marker in the target image, thereby improving the accuracy of the target answer.

This disclosure provides a model training method, apparatus, and device for question response, which can be applied to the field of data processing technology. The method includes: performing a text generation task on an initial sample question based on prompt options to obtain an inference chain corresponding to the prompt options; scoring the inference chain to obtain a scoring result, where the scoring result represents the semantic correlation between the sample question and the sample answer in the inference chain, and the semantic similarity between the sample answer and preset sensitive words; training a scoring model based on the scoring result and the sample scoring result obtained by processing the inference chain using the scoring model to obtain a target scoring model; training a policy model based on the sample question to obtain a target policy model, where the policy model is used to process the sample question and output sample response results, and the target scoring model is used to score the sample response results to obtain sample scoring results; and training the policy model based on the sample scoring results to obtain the target policy model.

The application discloses a kind of fusion multi-source rice product and adaptive identification feature's rice sample automatic extraction method and system, method includes: fusion multidimensional natural environment factor constructs rice sample feature natural zoning unit;Based on rice sample feature natural zoning unit, determine rice initial sample;For rice initial sample point, based on Sentinel-2 optical and Sentinel-1 SAR image data, calculate the optical and SAR image data features of rice key growth period, construct the multi-source time series feature space of rice whole growth period;Based on the multi-source time series feature space of rice whole growth period, construct multidimensional rice identification feature;Based on multidimensional rice identification feature, extract rice sample.

This disclosure relates to a method for solving the fatigue failure probability function of a turbine shaft based on extended dimensionless integration. The method includes: standardizing the input variables under a first distribution parameter and determining the fatigue failure limit state function; obtaining multiple unit direction vectors in the standard normal space and determining a first limit state surface based on the fatigue failure limit state function; filtering the unit direction vectors in the standard normal space corresponding to the first distribution parameter to obtain initial samples and effective unit direction vectors; transforming the initial samples on the first limit state surface to a second limit state surface in the standard normal space corresponding to the second distribution parameter; using the information from the transformed samples, determining the target distance between the origin of the coordinate system in the standard normal space along the effective unit direction vector and the second limit state surface using an interpolation strategy; and using dimensionless integration to determine the failure probability function value corresponding to the second distribution parameter based on the target distance.

The application provides a method for preparing a cobalt sulfide electron probe standard sample by nanogrinding and high-pressure synthesis technology, which takes arsenopyrite monomineral powder as a matrix, adds metal cobalt powder into the matrix, ensures that the cobalt content in an initial sample composed of the arsenopyrite monomineral powder and the metal cobalt powder is greater than or equal to 0.2 wt%, selects an ethanol dispersant, obtains sample powder with nanoscale size by mechanical ball milling with the size of grinding balls changed, applies axial load to the sample powder, cold-presses the sample powder into a consolidated sample, and obtains a cobalt-containing arsenopyrite electron probe standard sample after polishing. The application solves the problem of low or uneven cobalt content in natural cobalt-containing sulfides, and successfully prepares a cobalt-containing arsenopyrite electron probe standard sample meeting the uniformity requirement under submicron spatial resolution by using nanogrinding technology and high-pressure synthesis sample preparation method.

This application discloses a method and system for developing a closed-loop laser stealth cutting process, relating to the field of laser cutting technology. The method includes: acquiring historical processing parameters from past cutting processes and using these parameters as initial sample data; iteratively training a preset surrogate model using the initial sample data to obtain a trained first surrogate model; determining a combination of required parameters based on the first surrogate model and a multi-objective optimization strategy, and iteratively updating the first surrogate model multiple times using this combination until the first surrogate model outputs optimal process parameters that satisfy the preset optimization objective; and performing laser cutting on the target material using the optimal process parameters. This application achieves the technical effect of adaptively adapting to various laser cutting conditions for different cutting materials, thereby improving processing quality and efficiency.

This application discloses an APU rotor assembly balancing method and system, belonging to the field of aerospace technology. It includes: determining initial sample points based on initial imbalance test data; optimizing the initial sample points to determine test results under recommended assembly angles; if the requirements are not met, constructing expanded sample points by combining current and historical test data; estimating the probability density of the expanded sample points to obtain updated sample points with the highest probability density; and iteratively optimizing based on the updated sample points until the balancing requirements are met. This application, by constructing expanded sample points and using probability density estimation to screen optimal parameters, achieves dynamic and accurate correction of component imbalance, effectively overcoming the shortcomings of existing technologies that rely on predicted quantities or blind trial and error, and improving parameter estimation accuracy and convergence speed.

This invention relates to the field of semiconductor technology, providing a FIB sample preparation method and equipment for CL testing, and a CL testing method, comprising: based on a staged FIB processing approach, sequentially performing a preliminary cross-sectional cutting and a cross-sectional trimming of the region to be characterized in the sample to be tested to obtain a target cross-section; wherein, the surface of the initial sample cross-section obtained after the preliminary cross-sectional cutting contains an amorphous layer and an ion-implanted layer; after the cross-sectional trimming, the amorphous layer and ion-implanted layer are removed, and the flatness and tilt angle of the target cross-section meet the CL testing objectives. Through staged processing, first rough processing followed by fine trimming, the amorphous layer and ion-implanted layer generated due to lattice damage caused by high-energy ion implantation during processing can affect the photoelectric properties of subsequent cross-sections. By setting corresponding trimming stages, a target cross-section with a flat surface and low damage to the sample surface, without affecting photoelectric properties, can be obtained, improving the adaptability and accuracy of CL testing.

Disclosed is a technology for training artificial intelligence models using noisy labeled samples. More particularly, a method by which a training apparatus according to an embodiment of the present specification trains artificial intelligence models includes: relabeling samples through the artificial intelligence models, and selecting samples to be used for training from among the relabeled samples as first samples; extracting a structural label for each of the samples based on a relationship between the sample and other samples; and calculating a loss based on the first samples and the structural label.

This invention provides a method for constructing a coronary heart disease risk assessment model based on chest X-ray images, comprising the following steps: S1, acquiring chest X-ray data as the original clinical dataset to construct an initial sample set; S2, preprocessing the chest X-ray images in the initial sample set to obtain standardized image data; S3, dividing the standardized image data into a training set, a validation set, and a test set; S4, constructing a risk assessment model with two parallel branches: a convolutional neural network and a visual transformer; S5, iteratively optimizing the hyperparameters of the convolutional neural network model based on the performance of the validation set during training to obtain an optimized risk assessment model; S6, evaluating the performance of the optimized risk assessment model using the test set to obtain the final model for coronary heart disease risk assessment. This invention provides an efficient, accurate, low-cost, and non-invasive auxiliary diagnostic solution for coronary heart disease, which helps improve the early screening and treatment of coronary heart disease.

The present application relates to the technical field of sheet metal stamping forming and process optimization, and provides a method for optimizing stamping process parameters of a thin-walled battery shell based on LNN prediction, which comprises the following steps: establishing a new energy thin-walled battery shell geometric model, selecting three process parameters with the greatest impact on forming quality as optimization variables, and selecting an index for evaluating forming quality; generating a plurality of experimental combinations in finite element simulation calculation by adopting BBD design test scheme, extracting the maximum thinning rate and the maximum thickening rate corresponding to each test to form an initial sample data set; constructing and training an improved LNN prediction model; taking the trained LNN prediction model as a fitness evaluation function of MOPSO algorithm, predicting a Pareto optimal solution set, and selecting a unique optimal solution. The present application can capture the complex coupling relationship and nonlinear effect among multiple process parameters, and effectively realize the collaborative optimization of contradictory quality objectives.

The application focuses on the field of agricultural informatization and remote sensing application technology, and particularly relates to a kind of alfalfa yield estimation method based on multispectral and physical constraint sample enhancement. The method first acquires unmanned aerial vehicle multispectral image data corresponding to the alfalfa growth period, and lays out field quadrats with equal area in the image coverage area, collects measured data of alfalfa hay yield, leaf area index, chlorophyll content and leaf equivalent water thickness, and constructs an initial sample set. Then, input each parameter data within a reasonable range, generate predicted multispectral samples that meet the physical consistency constraint using the radiation transfer model, to expand the sample space. The measured samples and predicted samples are fused to construct an alfalfa hay yield estimation training data set. Based on the training data set, a mapping model between unmanned aerial vehicle multispectral features and alfalfa hay yield is established, and the alfalfa hay yield is estimated through machine learning.

The application relates to the field of droplet microfluidic digital analysis and detection, in particular to application of a crescent microsphere in digital polymerase chain reaction. The application provides a crescent microsphere, the inner diameter and the outer diameter of which can be adjusted according to specific experimental requirements. The application uses a method of double aqueous phase to prepare a special crescent-shaped hydrogel structure. The cavity of the crescent-shaped structure can accommodate a certain volume of aqueous phase, thereby reducing the repulsion effect of microsphere-water molecules in the pure spherical hydrogel microsphere, so that the liquid droplet formed based on the crescent structure has a larger polymerase chain reaction space. When the crescent microsphere is used in polymerase chain reaction, fluorescent intercalating dyes can mark positive droplets with targets and negative droplets without targets, the proportion of the positive droplets is counted, and in combination with Poisson distribution, more accurate digital quantification of the initial sample can be obtained.

The application discloses a sample model training method and system based on data driving, relates to the technical field of model training, and comprises the following steps: obtaining an initial sample set and inputting the initial sample set and a noise vector into a generator together to generate an initial prediction sample; constructing a composite feature tensor based on the initial sample, monitoring the difference between the prediction sample and a reference label in a feature space of a discriminator and the change trajectory of a total loss of the generator, and outputting a training termination condition tensor; feeding back the tensor to the generator as a condition constraint to update the prediction sample, and inputting the prediction sample and the initial sample set into the discriminator together to perform adversarial training, wherein the total loss of the generator is a weighted sum of an adversarial loss and a regularization loss, and network parameters are updated in reverse propagation according to the total loss; in the process of continuous updating of the network parameters, feature resampling is triggered to dynamically correct training data, and a final generator network is output when the difference between two consecutive rounds of training reaches a convergence threshold, so that the method realizes closed-loop correction of adaptive determination of training convergence.

This invention provides a method for predicting and optimizing the sound insulation performance of rail vehicles. The method includes: constructing and maintaining a unified database to centrally and standardizedly manage sound insulation material parameters and sound insulation sample data of the vehicle body structure; performing data augmentation based on initial samples, combining feature weighting processing with material layer location, and then determining feature weights through the mRMR algorithm and exponential decay weight strategy to finally train a sound insulation performance prediction model; using the database and the sound insulation performance prediction model to predict and evaluate the sound insulation performance of user-defined material combination schemes; and automatically searching for the optimal sound insulation scheme based on the database and the sound insulation performance prediction model, using an adaptively selected optimization algorithm, according to the scheme and constraints set by the user. This invention solves the problems of scattered data management, limited model prediction accuracy, reliance on experience in scheme design, and low optimization efficiency in traditional sound insulation design.

The application provides a full-size subsonic wind tunnel loop aerodynamic profile multi-objective optimization method based on steady CFD, and belongs to the technical field of wind tunnels.The application reduces the initial geometric parameters to a sensitive parameter subspace through principal component analysis, generates initial sample points by using Latin hypercube sampling, extracts response values by executing steady Reynolds average Navier-Stokes equation solving, constructs a Kriging surrogate model, performs multi-start parallel search based on an expected improvement function, extracts a separation zone length and a backflow zone height to input an unsteady correction factor model to calculate a correction coefficient for correcting a diffusion section pressure recovery coefficient, searches a Pareto front solution set by using a non-dominated sorting genetic algorithm, selects an optimal parameter combination through a comprehensive performance index, and reconstructs the wind tunnel loop aerodynamic profile through a hierarchical parameterization method and performs verification calculation, so that the problem that the steady calculation framework cannot accurately predict the separation flow and the optimization result deviates from the true performance is solved.

The present disclosure provides a training method of a deep learning model, relates to the technical field of artificial intelligence, in particular to the technical field of deep learning and the technical field of industrial big data. The specific implementation scheme is: generating enhanced sample data according to preset cooling knowledge information and initial sample data, wherein the initial sample data includes an initial value of at least one candidate cooling parameter of a sample material, the enhanced sample data includes an enhanced value of the at least one candidate cooling parameter, and the preset cooling knowledge information is used to indicate the relationship between the candidate cooling parameter and the target cooling temperature of the sample material; inputting the enhanced sample data into the deep learning model to obtain an enhanced sample output value of the target cooling temperature; obtaining an enhanced sample loss according to the enhanced sample output value of the target cooling temperature and a label value of the target cooling temperature; and training the deep learning model according to the enhanced sample loss. The present disclosure also provides a material data processing method and device, an electronic device and a storage medium.

The application discloses a kind of full-automatic process simulation method for product design, multiple-factor influence, multiple-target optimization, comprising:1, the data interaction between Creo Parametric and Ansys is established;2, parameterized model is established in Creo Parametric;3, parameterized model is imported into Ansys, meshing, case setting and result processing are carried out, and visual result is obtained;4, based on visual result, DOE experimental design is carried out in Desigin-Expert, and initial sample space is generated;5, initial sample space is imported into the parameter set of Ansys, parameterized simulation is carried out, and initial sample space target value is obtained;6, initial sample space target value is imported into Design-Expert, and target function between multiple factors and multiple targets is fitted;7, based on multiple-target optimization algorithm and weight setting, multiple-target optimization is carried out, and multiple-target optimization result is obtained;8, multiple-target optimization result is returned to Ansys and simulation verification is carried out.The application realizes the automation and process simulation under different structure parameters, and saves simulation time.

This invention discloses a reservoir injection-production optimization method based on a kernel-scale dynamically updated surrogate model, belonging to the field of reservoir production optimization technology. The invention first constructs a reservoir numerical simulation model, uses Latin hypercube sampling to obtain initial sample points and their target values ​​to establish a database, calculates the candidate scale pool of the surrogate model in the current iteration stage based on the spatial distribution characteristics of the samples in the database, and locally adaptively adjusts the candidate scale range according to changes in the search state during the evolution process. An elite training set is constructed by selecting elite samples, and a kernel-scale dynamically updated surrogate model is trained based on this. Candidate solutions are generated using a differential evolution algorithm, and the candidate scales are evaluated using ranking-weighted cross-validation error. The optimal kernel scale parameters for the current stage are selected and periodically updated. The updated surrogate model is used to predict, filter, and optimize candidate solutions to obtain the optimal injection-production regime, thereby improving the optimization efficiency and accuracy of injection-production regimes in complex reservoirs.

The application discloses an Android malicious program detection method based on a small sample learning model, and comprises the following steps: step 1, collecting initial sample data and performing preliminary preprocessing to obtain an original Dalvik operation code sequence of the initial sample data; step 2, performing Smali embedding representation on the original Dalvik operation code sequence of the initial sample data obtained in step 1; step 3, obtaining newly generated sample data; step 4, outputting a three-order fusion feature tensor with Android application malicious code semantic information; and step 5, providing a classification result of detecting whether an Android application is a malicious program. According to the method, the sample data enhancement is performed on the initial small sample data, the sample category imbalance problem is alleviated, and the problems of missing long sequence context association, high artificial feature engineering cost and high false detection rate are solved.

This invention belongs to the field of pharmaceutical quality testing technology. It discloses a high-throughput detection method for trace components in pharmaceuticals based on enzyme-specific reactions. The method includes: obtaining the initial sample addition timestamps of each well in a microplate that triggers the enzyme-specific reaction, as well as the dead time of mechanical flow; scanning and analyzing the high-frequency fluorescence signals in each well within an observation window after the dead time ends to obtain the fluorescence signal evolution matrix of the microplate; judging and labeling the state of each well by analyzing the obtained nonlinear distortion trend; reconstructing and extrapolating the dead time of the rapidly saturated wells to obtain the concentration values ​​of trace components exceeding the standard; analyzing the subthreshold fluctuations of the latent wells to obtain the subthreshold concentration values ​​of trace components; analyzing the photoelectric information of the normal wells to obtain the concentration values ​​of normal trace components; and combining the concentration values ​​of trace components exceeding the standard and the subthreshold concentration values ​​to form a pharmaceutical trace component detection report, thus reducing the blind zone of the detection range.

This invention relates to a method and apparatus for tuning control parameters of magnetic levitation bearings, a storage medium, and an electronic device, belonging to the technical field of magnetic levitation control parameters. The tuning method includes: S1, constructing a mathematical model of a PID parameter optimization problem; S2, generating a Bayesian optimization sample set, fitting a Gaussian process surrogate model, and globally exploring to generate a PID parameter subspace; S3, executing a dung beetle optimization algorithm within the subspace to obtain the optimal point (x) within the subspace. db y db S4, add the optimal point to the Bayesian optimization sample set and update the Gaussian process model; S5, repeat steps S2 to S4 until the condition is met, and output the optimal PID parameter x. db * By combining the global exploration capability of Bayesian algorithms with the local development capability of the dung beetle algorithm, the Bayesian optimization reduces the dependence on the initial sample size, while the dung beetle algorithm accelerates convergence within the subspace, thus balancing efficiency and accuracy.

Provided in the embodiments of the present disclosure are a data construction method, a code question-answering method, a task platform and a code question-answering system. The data construction method comprises: acquiring a first code snippet and an initial sample question for the first code snippet; using a text processing model to rewrite the initial sample question by using directory structure information of a sample code repository in a code library as context information, so as to obtain a sample question; on the basis of the first code snippet and the sample question, recalling a related second code snippet from the code library; and using the text processing model to generate, by using the second code snippet as context information, a sample answer corresponding to the sample question, wherein the sample question and the sample answer are sample data used for training a code question-answering model, and the code question-answering model is applicable to the code repository. The data construction efficiency is improved, and the data construction cost is reduced; and a model obtained by means of training is adapted to a code repository, thereby improving the user experience.

The application provides an air traffic control radar modeling method, comprising the following steps: selecting initial sample points, and obtaining the flyable range of each initial sample point by designing a simulation system for an air traffic control radar; constructing a training sample set; reducing the data dimension by using a principal component dimension reduction method and reconstructing the kernel function of a Kriging model; selecting different correlation functions and regression functions for combination, obtaining different Kriging model modeling schemes; solving the weight of each Kriging model modeling scheme; obtaining the output result of each Kriging model modeling scheme; constructing a weighted average aircraft flight range prediction model based on the output result of each Kriging model modeling scheme and the weight of each Kriging model modeling scheme; and the model can quickly calculate the reasonable flight range of an aircraft according to weather and air resistance information. The application improves the efficiency and accuracy of air traffic control radar modeling.