29 results about "Model testing" patented technology

A robustness measurement method for LeNet-5 networks based on adversarial boundary constraints is presented, relating to the field of deep learning model testing. The main steps include: for each training sample, dynamically generating adversarial examples based on the model's current state during training iterations; constructing a composite loss function based on standard cross-entropy loss and dynamic boundary constraint loss; performing end-to-end training on all parameters of the LeNet-5 network; and using the overall approximate robustness boundary as a measure of model robustness after training. This method effectively improves the resistance of the LeNet-5 model to fast gradient sign-based adversarial attacks without altering the basic structure of the LeNet-5 network by designing a new loss function.

This invention relates to the field of natural gas development and discloses a method, apparatus, storage medium, and processor for predicting the production of bottom-water gas reservoirs. The method includes: training a model based on a training set of production data from the bottom-water gas reservoir to obtain a first long short-term memory (LSTM) neural network model; optimizing hyperparameters based on a validation set to obtain a second LTM neural network model and a corresponding first error; testing the model based on a test set to obtain a corresponding second error; determining the generalization degree based on the first and second errors, and using the second LTM neural network model with a generalization degree meeting preset requirements as the target bottom-water gas reservoir production prediction model; and predicting the production of the bottom-water gas reservoir based on the target bottom-water gas reservoir production prediction model. This method captures long-term dependencies in time-series data using a LTM neural network, which can significantly improve the accuracy of bottom-water gas reservoir production prediction, thereby providing reliable support for oil and gas exploration and production decisions.

The application discloses a kind of based on DeepSeek's software and hardware collaborative optimization method and system, belong to computer software and hardware optimization technical field.The method is by collecting hardware and software related data in computer system and pre-processing, constructs DeepSeek comprehensive analysis model to learn the performance influence law of software and hardware interaction relationship, generates diversified test scene based on user demand and system characteristics and is automatically executed, according to model output formulates optimization strategy and then iteratively optimizes.The system includes data acquisition, pre-processing, DeepSeek analysis model, test scene generation and execution and optimization strategy formulation and implementation module, realize software and hardware collaborative optimization, break through independent optimization limit, improve test and optimization efficiency, reduce resource waste, enhance the adaptability to different computer systems, guarantee system running stability and reliability.

The present application relates to the technical field of sensor testing, and particularly relates to a GIS sensor real model testing device and method, through the setting of a GIS tooling, a measurement control system, a temperature and humidity generator, an SF6 gas mixing system, a vibration generator and an impulse voltage generator, a real running electromagnetic field environment can be simulated, various typical environmental disturbances can be applied, through multi-system linkage, the accuracy and sensitivity of the sensor in actual operation can be effectively verified, at the same time, reliable basis can be provided for the research on the cumulative influence of the strong electromagnetic environment, long-period vibration and instantaneous impulse voltage of the primary equipment of a substation on the performance of the sensor, the blank of reliability evaluation of long-term operation of the sensor is filled, and the problem of poor reliability of sensor test results caused by the mismatch between sensor testing and actual operation conditions in the prior art is solved.

The application provides a fuzzy test method for multimodal large model application. According to a preset test requirement, a corresponding test mode is selected as a target test mode, and under the target test mode, a test input content is constructed based on a preset prompt word library to test the application model under test. Meanwhile, based on a 3D label mapping algorithm, dimension label processing is performed on the test output content to obtain target test output content including a label dimension. Finally, according to the label dimension of the target test output content, the risk level of the application model under test is determined, so that model test of the application model under test is realized.

This application discloses a large-scale model testing method and related devices, electronic devices, and storage media. The large-scale model testing method includes: generating task objects based on test cases of the target large-scale model; controlling the runner to execute task objects in the task queue; in response to the runner failing to obtain a response after exceeding a target time threshold, controlling the runner to call a multimodal large-scale model to identify the page image of the browser page, obtain the target anomaly category, and controlling the runner to execute the target handling action matching the target anomaly category; and controlling the runner to re-execute the currently unsuccessful task object. This solution can minimize data crosstalk between different tasks and improve self-healing capabilities under abnormal scenarios while achieving concurrent testing of large-scale models with multiple tasks.

The application discloses a face soft tissue defect point cloud completion method based on a PF-Net, and belongs to the technical field of face soft tissue defect completion. The method comprises the following steps: S1, a data preparation stage: a data enhancement algorithm is designed to obtain a paired data set of incomplete input and defect labels; S2, a model construction stage: a dynamic graph convolution and a boundary attention mechanism are introduced into an encoder to optimize a point cloud completion network model PF-Net; S3, a model training stage: a training set is input into the PF-Net model for training; S4, a model testing stage: a test set is input into a generator for testing; and S5, a model inference stage: for clinical face soft defect point cloud data, multi-scale input data are acquired and completion is performed. The application realizes autonomous simulation design of face soft tissue defects, solves technical barriers in existing digital prosthetic technology, such as dependence on healthy side mirror image data or standard organ models, and solves the problem of unstable repair effect caused by dependence on doctor experience and aesthetic literacy.

The method comprises the following steps: constructing a large model evaluation workflow; obtaining a large model test set corresponding to a target evaluation field through a test set acquisition node; determining a test dimension meeting a preset evaluation quality index according to the large model test set through a test dimension determination node; sampling evaluation samples in the large model test set according to the test dimension to obtain a target test set through a target test set determination node; generating an evaluation script through an evaluation code generation node; and evaluating a to-be-evaluated large model by using the target test set to obtain an evaluation result. The method provided by the present disclosure selects a corresponding target test set based on a target evaluation field, thereby avoiding the problem of insufficient evaluation of the to-be-evaluated large model; reduces the waste of computing power and improves the evaluation efficiency; and evaluates the to-be-evaluated large model through multiple test dimensions to obtain an evaluation result with higher credibility.

The application provides a model testing method and device, electronic equipment and storage medium, comprising: acquiring time sequence monitoring data of a cloud platform; the time sequence monitoring data is used to represent the infrastructure state and service performance state of the cloud platform; according to the operation and maintenance business logic of the cloud platform, the time sequence monitoring data is labeled to obtain a first evaluation data set; the first evaluation data set is a full-amount benchmark test set covering a complete monitoring period; according to a predefined scene rule, the time sequence monitoring data is labeled to obtain a second evaluation data set; the second evaluation data set is a special test set for at least one specific operation and maintenance scene; the first evaluation data set and the second evaluation data set are respectively input into a time sequence prediction model to be evaluated to obtain a full-amount prediction result sequence and a sceneized prediction result sequence; and a test result is output according to the time sequence monitoring data, the full-amount prediction result sequence and the sceneized prediction result sequence.

The application discloses a power distribution network fault detection method based on a lightweight neural network, which comprises data acquisition, data preprocessing, data set division, neural network model construction, model training, model testing and fault detection; through integration of a pyramid lightweight extraction module PPD, a multi-scale feature splicing and fusion module MSCat and a cross-scale dynamic enhancement module DSE, an efficient feature extraction and fusion mechanism is constructed, the perception and recognition ability of multi-scale fault features is effectively improved while the model parameter quantity and the calculation burden are significantly reduced, and thus a solution is provided for realizing accurate, real-time and end-side deployable intelligent fault detection. The application solves the problems of high missing detection rate, limited recognition accuracy, high model complexity and difficulty in deployment in actual power distribution terminals with limited computing resources in the FTU fault detection method.

This invention discloses a calibration method for visual language models adapted during testing based on a single preparation. It utilizes a generative model to offline synthesize a small number of images for each category in the target domain, extracts features, and divides them into a pre-filled set and a calibration set. The pre-filled set is used to initialize positive and negative sample caches, achieving a "warm start." A lightweight probe network is trained based on the calibration set, which dynamically predicts the optimal fusion weights based on the original prediction scores and cached retrieval scores. During the testing phase, the network generates weights in real time and adaptively weights and fuses multiple predictions to obtain the calibrated prediction results. This invention requires only a one-time offline preparation, with extremely low online inference overhead. While maintaining efficient inference, it significantly reduces calibration errors, improving the reliability and security of model predictions, making it particularly suitable for real-time applications with strict confidence requirements.

The application relates to an unbalanced small sample fault diagnosis method based on preference calibration meta learning, and belongs to the technical field of fault diagnosis. The method comprises the following steps: obtaining mechanical equipment operation data and dividing the operation data into source domain samples and target domain samples; a fault diagnosis model is constructed based on meta learning, and the model comprises a multi-scale attention enhanced feature extractor and a Stiefel linear layer connected in sequence; in the model training stage, meta tasks are generated based on a source domain support set and a query set, and a double-layer optimization mechanism adopting an integrated preference calibration Riemann optimization strategy is used to update model parameters; in the parameter updating process, a dynamic class weighting matrix is updated through an adaptive preference updating mechanism, so that the focus of the model is shifted to learning insufficient classes, and effective learning of the model on all fault classes is ensured; in the model testing stage, the model is tested based on meta tasks generated based on a target domain support set and a query set. The application can improve the accuracy of mechanical equipment fault diagnosis.

This application provides a method and apparatus for filtering load testing data for model inference servers, belonging to the field of load testing data filtering technology. The method provided in this application involves: acquiring multiple load testing data sets for the same test item from a model inference server to form a load testing dataset; calculating the first-word latency of each load testing data set in the dataset; calculating the coefficient of variation and relative range of the dataset based on the first-word latency; performing anomaly judgment on each load testing data set based on the coefficient of variation and relative range, removing abnormal data to obtain a candidate dataset; and selecting the optimal load testing data from the candidate dataset based on the throughput and latency of the load testing data. The method and apparatus provided in this application for filtering load testing data for model inference servers are used to improve the stability of model test item samples and the reliability of results.

This invention discloses an emotion recognition method based on EEG and eye movement signals using modality generation, comprising the following steps: manual feature extraction; extraction of EEG spatial features; extraction of EEG temporal features; extraction of eye movement spatiotemporal features; feature concatenation and modality alignment; modality fusion; pre-training; and model testing. This invention introduces a modality generation pre-training task. Modality generation aims to generate a label for another modality given one modality as a condition. With the help of the pre-training mechanism, the model can learn robust and universal feature representations that adapt to various input conditions, effectively solving the accuracy decline problem caused by the missing modalities in multimodal emotion recognition, and significantly enhancing the model's adaptability to both modality-complete and modality-missing scenarios.

This invention relates to a simulation test system for intelligent contour-following excavation of large-angle inclined non-circular tunnels, belonging to the field of underground engineering model testing. It includes a support platform, an excavation device, a dust collection mechanism, and a cooling mechanism. The excavation device is mounted on the support platform, which is used to adjust the height and excavation angle of the device. The dust collection mechanism, installed on the excavation device, is used to absorb dust generated during the excavation process. The cooling mechanism, also installed on the excavation device, is used to cool the device. This invention allows for a wide range of adjustments to the height and inclination angle of the non-circular tunnel excavation device using a large-angle inclined support platform, enabling tunnel excavation at different heights and angles. The non-circular tunnel excavation device can meet the simulation excavation needs of tunnels with different shapes.

Embodiments of the present disclosure provide a code processing model training method, a code processing model testing method, and a code task processing method. The code processing model training method comprises: acquiring a sample code document and sample question-answer code; extracting sample completion code and sample context code of the sample completion code from the sample code document; converting the sample context code into a question-answer format to obtain sample instruction code; and training an initial processing model using the sample question-answer code, the sample completion code, and the sample instruction code to obtain a code processing model. A model is trained using sample question-answer code, sample completion code, and sample instruction code, enabling a code processing model to have both code question-answering and completion capabilities. By converting sample context code into a question-answer format, the processing capability of a code processing model is improved.

The application discloses a comprehensive evaluation test method and system based on large model understanding ability, and relates to the technical field of computers. First, rich and varied texts and multi-modal data are acquired as a test set, the text and multi-modal associated understanding are taken into account, and the large model capability is comprehensively investigated. Second, preset comprehensive and detailed evaluation indexes can determine the advantages and disadvantages of the model and propose optimization suggestions, thereby helping targeted optimization. Third, the test data set is regularly updated and expanded, the timeliness and comprehensiveness of the evaluation are maintained, the evaluation result is more in line with actual application requirements, and the problems that the evaluation data set of the large model test method in the prior art is not rich in theme, type and difficulty, the evaluation method mainly focuses on text understanding and ignores multi-modal association, the evaluation indexes are not comprehensive and detailed enough, and there is a lack of optimization suggestions for the advantages and disadvantages of the model after the evaluation, resulting in one-sided and inaccurate evaluation results, which is not conducive to model improvement.

A reliability prediction method for an image classification neural network model includes: a reliability model is trained; and the trained reliability model predicts a reliability of the image classification neural network model. According to training and testing of the image classification neural network model, input features of the reliability model include a model training factor and a model testing factor. The model training factor characterizes data and model factors affecting the reliability of the image classification neural network model. The model testing factor characterizes a test sufficiency of the image classification neural network model. An output of the reliability model is a reliability prediction result of the image classification neural network model.

The application discloses a kind of intelligent calculation cluster test method, device, equipment, medium and computer program product, the method includes: the initialization configuration of intelligent calculation cluster is carried out, each test link is executed in turn, and each test link is optimized in the testing process;Wherein, test link includes single machine comprehensive test, single machine model test, cluster communication test, cluster performance test and cluster long stable test;According to the target configuration of each test link after optimization, automatically arrange all test links, form test task sequence;In response to acceptance test request, execute test task sequence, and automatically generate acceptance test report according to the intermediate result of each test link.The application can more effectively locate problem server by integrating multiple incremental test links to test and optimize intelligent calculation cluster, thereby improving the accuracy and efficiency of fault location, and can arrange tasks, generate test report according to the latest test results, effectively improve test efficiency.

A computer-implemented method includes generating a candidate machine learning model. The candidate machine learning model is configured to generate probability scores for members. The method includes training the candidate machine learning model using an initial data set, testing the candidate machine learning model to generate performance metrics indicative of a bias value correlated to the candidate machine learning model, and determining whether the performance metrics are below a threshold. In response to determining that the performance metrics are below the threshold the method includes generating a proxy feature based on the bias value, adding the proxy feature to the candidate machine learning model, calculating weights based on the bias value, updating the initial data set using the weights to generate an updated data set, and retraining the candidate machine learning model using the updated data set.

A model testing method includes: a target device determines an auxiliary test model used for testing a first model, the auxiliary test model is determined based on a first condition, and the first condition is associated with one or more of: a training mode of the first model, an acquisition mode of the first model, a training mode of a peer model of the first model, an acquisition mode of the peer model of the first model, whether the first model is obtainable, whether the peer model of the first model is obtainable, whether the first model has been successfully deployed, whether the peer model of the first model has been successfully deployed, whether a reference model has been successfully deployed, whether a peer model of the reference model has been successfully deployed, and whether a test device supports the peer model of the first model.

The application discloses a software model testing method and device, a storage medium and an electronic device, and relates to the technical field of software testing. The software model testing method comprises the following steps: determining a target test model in a target software system; acquiring a model test case matched with the target test model; calling the target test model according to the model test case, and obtaining target test data output by the target test model; and marking a test result of the model test case according to the target test data. The technical scheme of the embodiment of the application can realize efficient and accurate automatic model testing, and provides reliable guarantee for the stability of a software system.

The application discloses a 3D target detection method based on a cylinder sequence attention and a hollow expansion convolution, and belongs to the technical field of target detection. The method comprises the following steps: step S1, data preprocessing; step S2, network model construction and model training: firstly, point cloud data used for training is prepared, and the data comprises point cloud scene data and target label data; then, a network model is constructed, and the hyperparameters of each module are set; then, the training data is sent into the network model for training, the model is optimized by reducing the loss function of the network during the whole training process, and thus the best network model weight is obtained; finally, the network model weight is saved; and step S3, model testing: the model effect is verified by using test set data. The application can realize accurate and real-time 3D target detection, effectively solves the problems of rough coding mode and coding information limitation, effectively increases the model receptive field while maintaining the simplicity and efficiency of the cylinder model.

The application provides a test method of a recognition model, acquires a test data set and a to-be-recognized model, the test data set includes test pictures and label data of subject targets, and the test pictures include the subject targets; for each test picture in the test data set, the subject target in the test picture is processed by the to-be-recognized model to obtain a recognition result of the test picture; based on a first array and a second array, the recognition result of the test picture is judged, and the correct number and the error number of the recognition result are recorded, the first array is used for recording first judgment data, the first judgment data is used for judging whether the recognition result is correct, the second array is used for recording second judgment data, and the second judgment data is used for judging whether the recognition result is incorrect; after the judgment and the recording of all the test pictures in the test data set are performed, the final correct number and the final error number are obtained; and based on the final correct number and the final error number, a test result of the to-be-recognized model is calculated.

This invention discloses a handwritten answer recognition method based on a visual language model, belonging to the field of image recognition and detection technology. The method includes: collecting handwritten answer image data and corresponding labeled text to construct a basic dataset; pre-training a visual language model based on the basic dataset, establishing semantic associations between image features and text features in vector space through a cross-modal alignment layer, and learning structured output expression logic; generating recognition results based on the pre-trained visual language model using a fully supervised fine-tuning mode and instruction-driven training method; incorporating supplementary difficult sample data into the training set based on recognition error cases generated during model testing, and repeatedly executing the fully supervised fine-tuning process until the expected recognition performance is achieved; and using the fine-tuned visual language model to complete the recognition of handwritten answers. This invention enables accurate output of structured question information in educational scenarios.

This invention relates to the field of artificial intelligence technology and discloses a method for constructing adaptive supervision signals for continuous model testing. The method includes: constructing a synthetic knowledge base; obtaining a set of test images at the current time step to construct a joint image set; performing style bridging on the synthetic images in the joint image set based on the test images to obtain stylized synthetic images; extracting features from the stylized synthetic images and performing statistical alignment based on the test images to obtain a target shallow feature map; extracting deep features from the target shallow feature map and the test images; and calculating a multi-dimensional loss based on the deep features as a supervision signal. This invention adapts the synthetic images to the current target domain through style injection and statistical alignment, and constructs accurate, comprehensive, and unbiased supervision signals by calculating multi-dimensional losses. This solves the problem of scarce and unreliable supervision signals in adaptive scenarios during continuous testing, and helps support forward facilitation for continuous model adaptation.

The present application relates to the field of Yolo video target detection analysis, and specifically provides a secondary determination fighting recognition method and device, which has the following steps: S1, data set preparation; S2, network model building; S3, network model training; S4, model testing and application. Compared with the prior art, the present application uses Yolov7 as a basic framework for recognition and classification processing. The use of image training set avoids the shortcomings of insufficient training video data set and difficult annotation. Since it is non-time sequence detection, the high algorithm operation complexity is avoided.

This invention discloses a fully unattended pool test system for a wind-wave integrated system, belonging to the technical field of pool model testing for offshore floating wind-wave multi-energy integrated devices. It includes a sensing layer, a decision-making layer, a control layer, and an execution layer. The sensing layer synchronously collects data on waves, motion states, equipment operation, and loads across all dimensions. The decision-making layer integrates automated control, intelligent PTO optimization, and graded safety protection functions. The control layer achieves sequential, synchronous closed-loop control of wave generation, PTO damping adjustment, and aerodynamic load application. The execution layer is equipped with multiple functional modules to complete wave generation, aerodynamic load application, and continuous stepless adjustment of PTO damping. Using this system, multi-device collaborative control is achieved, eliminating manual dependence and enabling autonomous, fully automated test operation. This effectively improves the coupling control accuracy and test data accuracy of the wind-wave integrated test, enhancing the operational safety and continuity of the test system.

The present application belongs to the technical field of hydrological forecast, and particularly relates to a karst region hydrological forecast method based on a variable-order generalized Nash confluence model, comprising: collecting hydrological and meteorological data of a research region; constructing a karst region hydrological model based on the variable-order generalized Nash confluence model; parameterizing the variable-order generalized Nash confluence model; model testing and precision evaluation. The variable-order generalized Nash confluence model proposed by the present application contains the recession process of initial water storage in karst regions on one hand, and adopts a variable-order structure to reflect the time-varying response of runoff intensity to the confluence process on the other hand, so that the confluence process in karst regions can be more truly reflected. In addition, since the variable-order generalized Nash confluence model can directly describe the water flow movement law in the complex medium of karst regions, it is not necessary to divide water sources, so that the model structure can be simplified, the model parameters can be reduced, and the uncertainty of the model can be reduced, which can provide an effective theoretical tool for high-precision flood forecast and hydrological simulation in karst regions.