272 results about "Model testing" patented technology

A system and method for multi-scale temporal attention processing in multimodal technology deep learning systems. This system processes time-series, textual, sentiment, and structured tabular data across three hierarchically-organized temporal streams—quarterly, weekly, and intraday levels—with bidirectional cross-temporal information flow. Scale-specific attention mechanisms are optimized for respective temporal granularities, while an adaptive controller dynamically weights each temporal level based on real-time market volatility indicators. A multi-scale fusion processor integrates attention-weighted representations to generate temporally unified representations preserving both short-term market dynamics and long-term trends. This approach enables superior forecasting and risk assessment by leveraging temporal correlations across multiple time scales while automatically adapting to changing market conditions. The system facilitates interpretable AI analysis through attention visualization and enables synthetic scenario generation for model testing.

A system and methods for processing diverse financial data types using a jointly trained Vector-quantized Variational Autoencoder (VQ-VAE) and neural upsampler. This system efficiently handles time-series, textual, sentiment, and structured tabular data through specialized encoding modules. A novel fusion module integrates these encodings, capturing cross-modal relationships via attention mechanisms and gated fusion units. The fused representation is compressed into a discrete latent space by the VQ-VAE encoder, then reconstructed and enhanced by the VQ-VAE decoder and neural upsampler, respectively. Joint training optimizes all components simultaneously, using a comprehensive loss function that balances reconstruction quality across modalities with upsampling performance. This approach enables superior data compression, reconstruction, and analysis, leveraging inter-modal correlations to improve financial forecasting, risk assessment, and decision-making. The system's ability to explore the latent space facilitates generation of new, synthetic financial scenarios for robust model testing and strategy development.

The invention discloses an out-of-focus identification and restoration method for a microscopic system. The method comprises the following steps: 1, preparing a microscopic out-of-focus image data set and carrying out degradation modeling; 2, constructing an out-of-focus parameter prediction network based on multi-label parameter reasoning and microscopic system priori knowledge and an image restoration network based on a generative adversarial network, performing independent training, and then performing merging training through a circulation system formed by mutual connection based on predicted point spread function convolution and deconvolution operation; 3, model fine tuning based on a specific system and a new data set; and 4, performing model testing, performing large-view image sliding window detection and collage fusion, and explicitly outputting defocus parameters of the system during image shooting. The method can be applied to out-of-focus fuzzy recognition and restoration of microscopic imaging of various systems, and is beneficial to the accuracy of functions such as particle counting and particle size statistics of the systems, thereby further promoting the application of the full-depth-of-view microscopic system in biomedical detection.

The invention discloses a feature fusion-based loaded coal and rock mass instability failure early warning method and system, and the method comprises the following steps: obtaining the information of an internal strain field and an external strain field through a coal and rock loading deformation failure test, and constructing a whole space-time sample data set; pre-training an auto-encoder for realizing feature extraction and constructing a multi-modal image; integrating depth features extracted from different source images by adopting a fusion strategy based on a self-attention mechanism so as to realize feature fusion; during training, inputting a multi-modal image in the data set into the neural network to obtain a strain field spatio-temporal evolution model; after the test is completed, the corresponding stage of the strain field measured by the sensor and the prediction result of the model are compared and verified, the relevant parameters of the model are adjusted according to the accuracy, and the fitting and generalization ability of the model is enhanced.

The invention discloses a cross-fault tunnel test simulation method under the combined action of ground vibration and fault movement, and belongs to the technical field of tunnel engineering model tests.The method comprises the following steps that firstly, the similarity relation of models is determined; 2, manufacturing a tunnel lining test model; 3, manufacturing a test model box; fourthly, surrounding rock is poured; 5, pressurizing the model box to simulate the tunnel burial depth; step 6, simulating fault dislocation; and step 7, applying seismic oscillation. In the invention, the designed horseshoe-shaped flexible model box replaces a rigid model box adopted by a conventional model test, the test cost is low, the test can be completed on one set of test system without designing several additional devices for cooperation, the operation process is simple, the feasibility is high, and the test efficiency is improved by considering the influence of ground stress aiming at the characteristic that the tunnel passes through the active fault. The deformation and failure evolution law of the tunnel under the combined action of seismic oscillation and fault movement is comprehensively analyzed, and the real site condition is better reflected.

The invention relates to safety compliance evaluation, in particular to a safety compliance evaluation system and method based on a multi-modal large model, and a functional module cluster, which are used for managing each functional module in the system, supporting cross-modal consistency detection and comparing answer consistency under different modals. The self-developed security system is responsible for execution of an assessment task, including resistance test enhancement and built-in multi-granularity data variation strategy, improves assessment robustness by simulating a real attack scene, realizes end-to-end automation from data generation, tested model test to risk assessment, supports configurable parameters and reproducible results, and has the advantages of high reliability and high reliability. The self-iteration and version management of the evaluation model are realized, and the self-iteration mechanism is deeply combined with a compliance label system, so that the risk identification capability under legal, ethical and industrial specifications can be continuously optimized; according to the technical scheme provided by the invention, the limitation that evaluation can be carried out only under a single mode and a single rule in the prior art can be effectively overcome.

The invention discloses a satellite orbit forecasting method based on deep learning physical constraint loss, and the method comprises the following steps: 1, carrying out the normalization preprocessing of input data, forming a training data set and a test data set, and constructing batch processing training data; and 2, performing dimension expansion on sample data points in each window in the batch processing data formed in the step 1, constructing a multi-dimensional feature space of the sample points, and forming a batch processing input data format capable of being introduced into the model. And 3, performing forward reasoning on the batch data formed in the step 2 by using a model, and obtaining a batch processing orbit prediction value output by the model at the next moment through a CNN lightweight spatial-temporal feature extraction module and a BiLSTM bidirectional time sequence neural network module. And 4, taking the track prediction value obtained in the step 3 and the truth value label in the training set obtained in the step 1 as input, calculating to obtain a loss value of a current training iteration batch through a multi-random learning loss module fusing physical constraints, and performing reverse updating of model parameters to complete model training. And step five, through the steps two to four, performing reasoning verification on the model by using the test set formed in the step one, and comparing with a truth value in the test set to obtain a model test result.

The invention provides a bidirectional recurrent neural network acoustic logging curve reconstruction method. The reconstruction method comprises the steps of S1, acquiring data and performing sequence segmentation; s2, carrying out normalization processing on the logging curve; s3, carrying out superposition networking and training on the bidirectional recurrent neural network structure block; s4, model testing and parameter storage; and S5, reading the model parameter file to reload the model, importing the to-be-reconstructed logging data, generating a prediction curve, and storing and exporting a result. A bidirectional recurrent neural network algorithm of artificial intelligence deep learning is adopted, bidirectional depth sequence information contained in a logging curve is effectively captured, the curve reconstruction accuracy is improved, unmeasured, missing and low-quality curves are reconstructed, multi-curve information is fused, and the sensitivity of a reconstructed acoustic curve is improved. The system has the advantages of being low in cost, high in efficiency and high in precision.

The invention provides a large model test case generation method, system and device based on multi-Agent collaboration. The method comprises the steps of obtaining product information of target product demand information; if the product demand information is incomplete, performing structured supplement on the extracted product demand information according to a slot filling technology to obtain complete product demand information; performing assembly according to the complete product demand information and the background information, and determining a first prompt word used for requesting the large model to generate a test case; and generating a first edition test case according to the first prompt word by using the plurality of agents, reviewing the first edition test case, generating an optimization suggestion, and fusing the first edition test case based on the optimization suggestion to generate a final test case. Through knowledge injection and structured requirements, the test case not only covers functional requirements, but also improves the accuracy of the test case; dynamic verification and correction of the use cases are achieved through cooperation of the multiple agents, the generation efficiency of the test use cases is improved, and the coverage rate is increased.

The invention discloses a remote sensing target detection large model construction method based on context feature deep learning, and the method comprises the following steps: S1, dividing an original data set according to a strategy, forming a training set, a verification set and a test set, and guaranteeing that each subset has representativeness in sample number, scene diversity and target distribution; s2, constructing a remote sensing target detection large model by adopting a deep learning method based on context features, and fusing multi-level features and introducing a query optimization mechanism; and S3, performing model testing and performance evaluation, selecting optimal model parameters, performing comprehensive verification on the test set, and evaluating the comprehensive performance of the model in the aspects of detection precision, speed and robustness. According to the method, the pre-trained DINOv2 backbone network, the context feature enhancement module and the intensive supervision strategy are introduced, so that rich context information can be captured in a multi-scale and complex scene, and the detection precision is greatly improved.

The invention discloses a model performance test method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. The theoretical maximum lexical throughput of a target large language model is calculated based on the video memory bandwidth of a graphics processor, the model parameter quantity, the byte number corresponding to the quantization precision and the video memory bandwidth utilization rate; meanwhile, the benchmark performance throughput is obtained, a theoretical corresponding first concurrency number is calculated in combination with the theoretical maximum lexical unit throughput and the concurrency competition loss coefficient, then the model test is executed based on the first concurrency number to obtain the actual maximum lexical unit throughput and a corresponding second concurrency number, and a model performance test result is generated. The problems that in the prior art, due to the fact that manual testing is conducted depending on manual intervention, a continuous approaching attempt mode is adopted, a reasonable test starting point is not deduced in combination with hardware core bottlenecks and key parameters, evaluation is time-consuming and labor-consuming, the result is prone to being affected by artificial factors, and accuracy and consistency are poor can be solved.

The embodiment of the invention provides a model testing method and device, equipment and a storage medium. The method comprises the steps that a plurality of local groups are created for a plurality of candidate defect types respectively; obtaining a plurality of target images, respectively inputting the plurality of target images into the defect detection model for detection to obtain a defect detection result of each target image, and for each target image, obtaining a defect detection result of a target defect type corresponding to the target image based on the first confidence interval and the target confidence of the target defect type corresponding to the target image; determining a first target group in each local group corresponding to the target defect type, and dividing the target image into the first target group; and for each candidate defect type, determining a first detection group in each local group corresponding to the candidate defect type, and determining a test result of the target model based on a relationship between the number of the target images corresponding to each first detection group and the total number of the target images. According to the embodiment of the invention, the stability and robustness of the model in a production field can be improved.

The invention relates to the technical field of model testing, in particular to a model testing method and device and electronic equipment, and is used for solving the problems of long test data combination time consumption and low coverage rate of a model testing mode in the related technology. The multiple features are classified, a core feature set, a common feature set and an edge feature set are obtained, and the model can be a deep learning model; generating a test set of the model according to a differentiated data generation strategy of feature importance on the basis of the rule set of the service type adapted to the model, the core feature set, the common feature set and the edge feature set, and performing model performance test on the model by using the test set to obtain a test report; therefore, the efficiency of model algorithm testing is improved.

The invention discloses a multi-dimensional large model test evaluation method and system, belongs to the technical field of artificial intelligence testing, integrates a basic model, platform capability and application performance to realize multi-dimensional large model test evaluation, and comprises the following steps: basic model capability evaluation, including sensitive lexicon + BERT semantic double-engine security detection and context dependence test chain; evaluating the capability of the model platform, including presetting LLM intelligent labels and double-person back-to-back data backflow; application performance evaluation, including character disturbance robustness test and LIME interpretability grading evaluation; large-model full-stack testing is achieved through a RESTful heterogeneous model nanotube interface, and localized compatibility adaptation verification is achieved based on GPU attenuation rate quantification. The method solves the problem of lack of evaluation of platform engineering capability, safety compliance and localization support in the prior art, realizes three-dimensional capability coupling evaluation, solves the problem of localization adaptation, and fills the blank of a large-model full-stack test technology.

The invention relates to the technical field of risk management, and discloses a financial AI large model-oriented multi-dimensional test intelligent risk management and control method, which comprises the steps of integrating channel test, data test, AI test and AI large model test by constructing a financial AI large model test special database; traditional financial sensitive fields and AI test specific sensitive fields are accurately identified based on data consanguinity analysis, and dynamic desensitization is realized in combination with a lightweight desensitization engine; calling a preset risk scene library, defining a risk index system of each scene, dynamically calculating a risk level through a weighted scoring algorithm, and carrying out modeling by using a graph database to identify hidden linkage risks; triggering a grading response strategy according to the risk grade, and dynamically optimizing resource allocation; and after the test is finished, a multi-dimensional risk thermodynamic diagram and a root cause analysis report are generated, and processing experience is fed back to the scene library optimization strategy. According to the invention, the efficiency and safety of risk management and control can be improved.

The invention discloses a ceramic firing performance prediction method based on a GA-BP neural network, and the method comprises the steps: data collection and preprocessing, gray correlation analysis and screening of key process parameters, construction of a GA-BP neural network model, model training and verification, prediction model testing, and firing performance index prediction. By means of unique intelligent algorithm fusion, the problems of high experience dependence, high experiment cost, long period, low efficiency, time consumption, energy consumption, high efficiency and the like of a method for judging the ceramic firing performance through artificial experience in actual ceramic firing production are effectively solved. The technical problem that multi-target collaborative optimization is difficult to realize at the same time in the prior art is solved, remarkable advantages are shown in the aspects of prediction precision, intelligent degree and the like, the requirements of intelligent, green and high-quality ceramic firing process optimization are met, and the method has wide application prospects and important practical value.

This disclosure describes a policy consistency system that determines when RBAC policies are inconsistently implemented across different devices, platforms, services, and environments. For example, the policy consistency system uses a model-based testing framework to identify cases where user requirement inputs produce different outputs when the policy is implemented in different environments. In some implementations, the policy consistency system utilizes symbolic abstraction models and satisfiability solver models to efficiently identify inconsistencies in the implementation of a policy.

The invention relates to the technical field of embedded artificial intelligence, and discloses an arc discharge detection model training method and system based on quantitative perception training, and the method comprises the steps: constructing an initial floating point precision neural network model for DC arc discharge fault detection, and defining a plurality of quantization functions for the model to convert floating point parameters; a FakeQuant pseudo quantization strategy is self-defined, a linear calculation layer in an original model is replaced with a quantifiable calculation layer, and a quantization operator is created through the quantifiable calculation layer; performing layer-by-layer differentiation quantitative perception training on the replaced model based on a training data set, and simulating a forward calculation behavior of a low-precision fixed-point type in actual reasoning while keeping floating-point precision back propagation through a quantization operator, so that the weight of the model is adaptive to a quantization error; and converting the trained model parameters into a target low-precision fixed-point type, and generating an end-side deployable arc discharge detection model through model testing and computing power evaluation, thereby realizing high-fidelity conversion from a floating-point model to a low-precision fixed-point model.

The invention belongs to the technical field of vehicle-mounted network intrusion detection, and relates to a BERT-based lightweight vehicle-mounted network intrusion detection method, which comprises the following steps: S1, a data preprocessing stage: converting a CAN sample with a label into a text sample; s2, a model training stage: inputting the text sample with the label into a VehileBERT series model for training, including setting a training hyper-parameter, and transmitting the processed data to the VehileBERT series model for training; and S3, a model test evaluation stage: performing reasoning test on the trained VehileBERT series model. The method has the advantages that the loss caused by light weight is balanced by adjusting the dynamic attention weight, and the VehileBERT model is finally designed to be suitable for vehicle-mounted network intrusion detection. And the real-time requirement of the vehicle-mounted network can be better met.

The invention discloses a tool and model testing method based on TTS, and belongs to the technical field of intelligent terminal testing. According to the method, the TTS technology is combined with natural language generation, environment simulation and a self-adaptive algorithm, a test text set containing sentence pattern variants and synonym replacement is generated through a natural language generation model, a parameter-adjustable TTS engine is used for synthesizing voice signals with acoustic and emotional characteristics, the voice signals and an environment impulse response function are subjected to convolution superposition of scene noise and then input into a terminal, and the voice signals are converted into voice signals with the acoustic and emotional characteristics. Mapping to a semantic vector space to calculate similarity, and finally generating an iterative test strategy according to the confusion matrix. According to the scheme, test text automatic generation, complex environment simulation and test strategy adaptive optimization are realized, the test efficiency and coverage are improved, the problem is accurately positioned, the multi-scene adaptability is enhanced, and a full-process intelligent test scheme is formed.

The invention discloses a multi-period non-stationary hydrological variable valuation method, and relates to the technical field of hydrological prediction. Comprising the steps of obtaining hydrological variables of a plurality of observation stations in a drainage basin at different time points as hydrological time sequences of the observation stations; constructing an experimental variation function according to the hydrological time sequence of each observation station; selecting a drift model through a trend surface analysis method and model inspection; constructing a multi-period non-stationary variable equation set comprising a Lagrange multiplier and a hydrological variable weight based on the experimental variation function and the drift model; expressing the multi-period non-stationary variable equation set as a matrix form, and calculating a Lagrange multiplier and a hydrological variable weight through a matrix inversion method; and constructing an estimation formula according to the calculation result of the Lagrange multiplier and the hydrological variable weight, and performing multi-period spatial estimation on the hydrological variable of the to-be-measured drainage basin through the estimation formula. Dependence on the number of observation stations can be reduced, and accurate estimation can be achieved only through a small amount of observation station data.

Provided in the embodiments of the present specification are a task processing method and apparatus, a question answering processing method and apparatus in a target domain, a domain task model test method and apparatus, and a computing device, a computer-readable storage medium and a computer program product. The task processing method comprises: acquiring task data of a target task, wherein the target task comprises at least one task stage, and the task stage comprises a thinking sub-task and a decision-making sub-task; inputting the task data into a domain task model, and executing the thinking sub-task, so as to obtain a target thinking result of the domain task model with respect to the task data; and inputting the target thinking result into the domain task model, and executing the decision-making sub-task, so as to obtain a target task result corresponding to the target task, wherein the domain task model is obtained by means of performing training on the basis of a test result of the thinking sub-task and a test result of the decision-making sub-task. By means of step-by-step interaction with a model, the thinking analysis capability and the task processing capability of the model with respect to a target task can be improved on the basis of a chain of thought of the model, thereby improving the accuracy of a target task result.

The technology described herein provides an automated software-testing platform that functions in an undefined action space. The technology described herein starts with an undefined action space but begins to learn about the action space through random exploration. Both the action taken during testing and the resulting state may be communicated to a centralized testing service. The technology described herein also mines the action telemetry data and state telemetry data to identify action patterns that produce a sought after result. Once a plurality of action patterns is identified and, at least, a partial model of the action space is built, the testing on the test machines may be split into random test mode, replay test mode, and a pioneering test mode.

The invention discloses a privacy protection and robustness test method and system for large model fine tuning, and belongs to the technical field of machine learning security. The method comprises the steps that a three-layer distributed architecture comprising an edge server, a cloud server and a plurality of edge clients is constructed, the edge clients distill local privacy data and cooperate with the edge server to train a global model, and a candidate detection sample set is formed; screening a sample set based on the potential feature deviation evaluation index, and sending the sample set to a cloud server for vulnerability detection to obtain an optimal backdoor detection candidate sample set; and multi-trigger parallel and progressive trigger sequence backdoor implantation is respectively used for scenes of single fine tuning and multiple fine tuning of the large model, an optimal backdoor detection candidate sample set is combined with a preset trigger to generate a backdoor test sample set, the backdoor test sample set is mixed with a clean data set, and then the robustness of the backdoor test sample set is tested through fine tuning of the large model. Large model fine tuning and robustness testing of privacy protection can be realized in a heterogeneous model cooperative training environment.

The invention discloses a manufacturing industry contract review method, device and equipment based on a large model, and a medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: converting a target paper contract image into a to-be-processed contract text through employing an optical character recognition technology, executing a preset data cleaning operation and a preset structured processing operation on the to-be-processed contract text based on a natural language processing technology to obtain a target contract text; dividing the target contract text into model training data, model verification data and model test data to train a preset contract evaluation large model, and optimizing the target parameters to obtain a target preset contract evaluation large model; and obtaining a target entity and a target entity relationship by using the target preset contract evaluation large model to determine a target contract element, and executing a preset compliance evaluation operation and a preset risk evaluation operation on the target contract element to obtain a target evaluation result so as to complete contract evaluation. In this way, automatic and intelligent contract review at a higher level is achieved.

The invention discloses a model training method and device, and the method comprises the steps: judging whether a target training model meets a model test condition or not according to a model test plan and the current training state of the target training model; if yes, copying the target training model to obtain at least one target test model; and performing model training on the target training model, performing asynchronous and parallel model testing on the target testing model, and judging whether the model training of the target training model is completed or not according to a model testing result. The resource utilization rate can be improved, and the model training efficiency is improved.

The invention relates to an air conditioner load prediction method based on a self-adaptive double-flow graph attention network, and belongs to the technical field of building energy conservation and intelligent control. The method comprises the following steps: collecting historical power and environmental data of an air conditioner, and after preprocessing and normalization, constructing an input sequence through a sliding window and dividing a data set according to time; a prediction model is constructed, and a causal graph learning module, a multi-scale graph structure learning module, a self-adaptive space-time attention module, an uncertainty quantization module and a self-adaptive sampling module are integrated; a training set and a joint loss function training model are adopted, and a load prediction result and uncertainty estimation are output through Monte Carlo Dropout during testing. According to the method, the dynamic causal relationship between variables and multi-scale space-time dependence can be adaptively learned, reliable uncertainty quantification is provided while the prediction precision is improved, and the method is suitable for intelligent regulation and control and energy efficiency optimization of the air conditioning system.

The invention relates to the field of mechanical fault identification, and discloses a rotating machine fault diagnosis method and system based on a multi-granularity knowledge supervision comparison optimization strategy, and the method comprises the steps: obtaining an original vibration signal, carrying out the window interception, and dividing a training data set and a test data set; performing normalization and data enhancement on each sample in the training data set twice to obtain a data enhancement sample for model training; samples in the test data set are only subjected to normalization processing and are used for model testing; building a granularity knowledge migration supervision comparison model; pre-training a feature extractor; training a granularity knowledge migration supervision comparison model by adopting the data enhancement sample according to a given granularity knowledge migration classification training step, a classification loss function and a back propagation optimization algorithm to obtain trained model parameters; and inputting a test sample into the trained granularity knowledge migration supervision comparison model to obtain a multi-fault classification and diagnosis result.

The embodiment of the invention discloses an automatic test method and system for an intelligent cockpit large model, and the method comprises the steps: uploading a to-be-tested case document, and generating a test input signal according to the to-be-tested case document; receiving the test input signal, executing a test task for the test input signal, and generating a test response result; collecting a test response result, wherein the test response result comprises a voice recognition result, a human-computer interaction interface screenshot and a system operation log; performing semantic comparison and scoring on the test response result through a judgment large model, and outputting a related index corresponding to the test response result; and generating a test result evaluation report according to the related indexes corresponding to the test response result, and outputting a typical error case list according to the test cases which do not reach the standard in the test result evaluation report. According to the method, the accuracy, the efficiency and the reproducibility of the intelligent cabin model test can be improved.

The invention relates to a keratoconus analysis system based on dynamic cornea textures. The keratoconus analysis system comprises an image acquisition unit, an image preprocessing unit, a feature extraction unit and a feature screening unit which are connected in sequence, the multi-time-point cornea dynamic image texture analysis method is provided, in order to ensure that the screened features have the optimal diagnostic capability, a feature set with the highest diagnostic value is screened through a recursive feature elimination method, and a foundation is laid for FFKC high-precision recognition. The machine learning classification system comprises a segmentation module, a model training module, a model testing module and a performance evaluation module, and is used for obtaining a machine learning classifier. The performance of the model is evaluated through an ROC curve and a confusion matrix, and high sensitivity and high specificity are ensured. A multi-modal recognition platform fusing dynamic textures and cornea biomechanical parameters is constructed, the platform sequentially comprises a data input module, a data processing module, a model module and an output module, and efficient FFKC recognition and auxiliary diagnosis are achieved.