Patents
Literature
Patsnap Eureka AI that helps you search prior art, draft patents, and assess FTO risks, powered by patent and scientific literature data.

74 results about "Confidence distribution" patented technology

In statistical inference, the concept of a confidence distribution (CD) has often been loosely referred to as a distribution function on the parameter space that can represent confidence intervals of all levels for a parameter of interest. Historically, it has typically been constructed by inverting the upper limits of lower sided confidence intervals of all levels, and it was also commonly associated with a fiducial interpretation (fiducial distribution), although it is a purely frequentist concept. A confidence distribution is NOT a probability distribution function of the parameter of interest, but may still be a function useful for making inferences.

Automatic driving decision-making system based on information entropy-Bayesian estimation

The invention provides an automatic driving decision-making system based on information entropy-Bayesian estimation. The automatic driving decision-making system comprises an environment sensing module, a hybrid quantification module and a decision-making optimization module. The environment sensing module is used for collecting and processing sensor data and completing target detection and classification; the hybrid quantization module performs reliability quantization on the recognition result based on information entropy and Bayesian uncertainty estimation, generates uncertainty evaluation results of different targets and scenes, and transmits the uncertainty evaluation results to the decision and control module; and the decision and control module dynamically adjusts the strategy of the automatic driving system according to the uncertainty measurement of each target. The automatic driving decision-making system based on information entropy-Bayesian estimation is used for quantifying the target recognition reliability in an automatic driving scene. The output confidence distribution of the target recognition model is evaluated by calculating the information entropy, and the uncertainty of the recognition model is estimated by using a Bayesian uncertainty estimation (such as Monte Carlo Dropout) method.
Owner:HARBIN INST OF TECH +1

High-noise data intelligent cleaning and credibility evaluation method based on uncertainty quantization

The invention relates to the technical field of data processing, in particular to a high-noise data intelligent cleaning and credibility evaluation method based on uncertainty quantization, which comprises the following steps: constructing a noise manifold model of a high-dimensional data space through a Riemannian geometry method, and extracting data intrinsic structure features and noise distribution trajectory tensor; according to the noise distribution trajectory tensor, constructing an anisotropic diffusion equation, and generating confidence distribution with data manifold constraint; the geometric invariance of a data set is kept while noise stripping is carried out in a data manifold tangent space, and noise stripping and structure keeping are achieved; calculating an intrinsic credibility index and generating a credibility evaluation map coupled with the data manifold structure; according to the method, the intrinsic characteristics of the data manifold are kept in the cleaning process, the deformation degree of the quantized data structure is calculated through the geodesic deviation, and it is ensured that the intrinsic geometrical relationship of the data is not affected by noise stripping.
Owner:HANGZHOU JINCHENG INFORMATION SECURITY TECH CO LTD

Text classification method based on large model

The invention discloses a large model-based text classification method, which comprises the following steps of: 1, setting a task semantic constraint rule, and writing a fixed cue word template; 2, sorting classification labels and adding semantic description information; 3, semantic similarity clustering and confidence allocation are executed, and a confidence multi-granularity label prefix tree is generated; 4, dynamic constraint autoregressive decoding is executed in combination with the confidence multi-granularity label prefix tree and the large language model, and a target classification label is output; 5, constructing a BERT auxiliary discrimination module to carry out confidence mask constraint; step 6, constructing a decoding temperature control parameter and a semantic bias item to perform semantic guidance and category distinguishing control; 7, performing uncertainty self-calibration on the autoregression decoding process of the large language model; and 8, constructing semantic consistency loss between the large language model and the BERT auxiliary discrimination module. According to the method, the stability and the classification efficiency of text classification in a small sample scene are improved.
Owner:KEXUN JIALIAN INFORMATION TECH CO LTD

Multi-relation extraction error propagation optimization method and device based on data collaborative enhancement

The invention discloses a multi-relation extraction error propagation optimization method and device based on data collaborative enhancement, and the method comprises the steps: generating extended training data through grammar recombination and adversarial samples, carrying out the positioning and weighted sampling of a low-frequency relation according to an entity pair relative position, and obtaining an upper relation data set; and training an upper relation classifier based on the upper relation data set, and performing prediction through the trained upper relation classifier to obtain a prediction result and confidence distribution thereof. According to the method, through text enhancement and a weighted sampling strategy for relative position positioning based on entities, the problem of unbalanced data distribution is effectively relieved, the modeling capability of the upper relation classifier for long-tail distribution is remarkably improved, and therefore systematic optimization of the upper relation classifier for low-frequency relation recognition accuracy is achieved.
Owner:WUHAN UNIV OF SCI & TECH

Uncertain knowledge graph reasoning method based on semi-supervised confidence distribution learning

The invention discloses an uncertain knowledge graph reasoning method based on semi-supervised confidence distribution learning. The method comprises the following steps: converting triple confidence in uncertain knowledge graph training data into confidence distribution; learning embedding of the uncertainty knowledge graph on the marked data and the pseudo-marked data generated by the pseudo-marked data generator at the same time by using a relation learner based on confidence distribution learning; generating a high-quality pseudo-confidence distribution label for the unmarked data by using a pseudo-marked data generator; iteratively training a relation learner based on confidence distribution learning and a pseudo-mark data generator by utilizing element self-training until the relation learner and the pseudo-mark data generator converge; and transmitting data to be complemented into the trained relationship learner based on confidence distribution learning for reasoning to realize uncertainty knowledge graph complementation. According to the method, the supervision information of a small amount of confidence or no confidence in the marked data can be captured, and the method is suitable for the scene of unbalanced confidence distribution of the triple of the uncertainty knowledge graph.
Owner:SOUTHEAST UNIV

Trusted evidence reasoning method, device and equipment for multi-source data fusion

ActiveCN120012944AMachine learningInference methodsEvidence reasoningSupport matrix
The invention relates to a credible evidence reasoning method, device and equipment for multi-source data fusion, a multi-source decision information system is constructed, in the multi-source decision information system, a support matrix is designed to quantify correlation among multi-source data attributes, and a relationship is evaluated by combining concepts of similarity and distance. Firstly, multi-source data of different scales are converted into confidence distribution, probability significance is given to the confidence distribution, and therefore generated evidence can be applied to multi-scale data fusion. Then, a credible evidence reasoning rule is constructed, and in the rule, the credibility of the evidence is determined by the weight and the reliability of the evidence; by integrating multi-source data samples of different scales, confidence distribution of results is derived. Therefore, the problem of difficulty in determining confidence distribution of evidences is solved.
Owner:NAT UNIV OF DEFENSE TECH

Analysis method and system of soil microbial community structure and medium

PendingCN120564854ABiostatisticsSequence analysisMicroorganismMultinomial logistic regression
The invention provides an analysis method and system for a soil microbial community structure and a medium, and relates to the technical field of ecological environment monitoring. According to the method, environmental parameters, namely pH, organic carbon, moisture and oxidation reduction potential, of a soil sample are collected and subjected to normalization processing, phylum abundance, dominant phylum information and Shannon index classification community structure categories are obtained in combination with high-throughput sequencing, the support probability of the environmental parameters for classification is calculated through a multi-term logistic regression model, and the classification result is obtained. A D-S evidence theory is utilized to fuse multi-source BPA (basic probability allocation), joint confidence distribution is output, a conflict factor threshold value is set, an artificial review prompt is triggered under a certain condition to avoid errors caused by environmental parameters, finally, to-be-detected soil parameters are input into a pre-training model, confidence distribution is output in real time, and the detection accuracy is improved. And judging the structure of the microbial community in combination with a threshold rule, and outputting health, risk, transition state or uncertainty.
Owner:黑龙江省农业科学院黑河分院

Cable fault diagnosis and dynamic maintenance method and system based on multi-domain feature fusion

The invention provides a cable fault diagnosis and dynamic maintenance method and system based on multi-domain feature fusion, and relates to the field of fault diagnosis. Time-varying weighted Clark transformation is adopted to obtain a current vector trajectory diagram based on cable current information; inputting the trajectory diagram into an improved wavelet CNN network to obtain a fault diagnosis result; the improved wavelet CNN network comprises a multi-stage feature extraction module and a feature fusion module which are connected in series, and the multi-stage feature extraction module is used for extracting low-frequency features stage by stage; a high-frequency feature enhancement branch is added behind the last-stage feature extraction module and is used for extracting high-frequency local distortion features; and obtaining a fault category and corresponding confidence distribution based on a fault diagnosis result, inputting the fault category and the corresponding confidence distribution into the large language model to obtain a semantic diagnosis report containing mechanism description, potential risks and maintenance suggestions, and performing fine adjustment on the large language model by using line identification parameters. Accuracy, robustness and interpretability of cable fault diagnosis are effectively improved, and powerful support is provided for intelligent operation and maintenance of cables.
Owner:JINING POWER SUPPLY CO OF STATE GRID SHANDONG ELECTRIC POWER CO

Priori mask optimization method based on similarity measurement

PendingCN120372305ANeural learning methodsAlgorithmFalse positive prediction
The invention discloses a priori mask optimization method based on similarity measurement. The overall process is as follows: foreground mask extraction based on a fusion prototype, background mask extraction based on a gating mechanism, and foreground mask optimization based on loop feedback. A trainable weighted fusion channel is established between a support prototype and a query prototype, the semantic deviation problem of traditional single prototype matching is solved, a background probability confidence coefficient matrix and cyclic consistency check are combined, cross-sample positioning of false positive prediction is achieved, temperature parameters are dynamically adjusted in the mask correction stage, and the accuracy of the temperature correction is improved. Compared with a fixed temperature strategy, the method is more suitable for confidence distribution characteristics of complex scenes; and the segmentation mask is optimized by fusing prototype similarity measurement and a cyclic feedback mechanism, so that false positive prediction is effectively reduced, and the distinguishing precision of the foreground and the background is improved.
Owner:SOUTH CHINA NORMAL UNIV

Hydraulic plunger pump fault online diagnosis method and system based on dynamic confidence distribution

The invention discloses a hydraulic plunger pump fault online diagnosis method and system based on dynamic confidence distribution, and relates to the technical field of intelligent detection, and the method comprises the steps: carrying out the feature extraction of the preprocessed real-time operation state data, combining the extracted features, and obtaining a total feature vector; acquiring a real-time signal of the hydraulic plunger pump, and constructing a dynamic confidence interval based on the real-time signal; if the signal value of the current real-time signal exceeds the dynamic confidence interval, calculating the confidence probability of the current real-time signal, and obtaining the joint confidence probability of all the signals based on the confidence probability; inputting the total feature vector into a pre-trained fault diagnosis model for prediction to obtain a prediction probability of a fault category; and obtaining a posterior probability of the fault category based on the prediction probability and the joint confidence probability, and determining a fault diagnosis result based on the posterior probability. Common faults such as abrasion, blockage and internal leakage of the hydraulic plunger pump can be accurately recognized, and the real-time performance and reliability of diagnosis can be improved.
Owner:LIAOCHENG UNIV

Cold storage energy-saving control method based on deep reinforcement learning

The invention relates to the technical field of refrigeration house energy-saving control, and provides a refrigeration house energy-saving control method based on deep reinforcement learning, which comprises the following steps: dividing a refrigeration house according to a multi-scale grid, collecting data in real time, weighting and aggregating according to spatial characteristics, calculating spatial semantic codes and confidence indexes of each grid, and obtaining a standardized and traceable state vector; real-time energy consumption, grid temperature space gradient, long-term exposure economic loss, event condition triggering and weighted summation form a reward function, and a deep strategy network is adopted to train parameters in an accumulated reward mode; constructing a grey box and residual dynamic prediction model, shrinking the original temperature and humidity constraint, solving the prediction model through scenario under the shrinkage constraint, and carrying out weighted fusion on the depth strategy obtained by training and the solution of the prediction model according to confidence; and calculating a statistical value in the sliding window, constructing confidence distribution of strategy return, triggering a hierarchical response according to a detection result, and inputting all triggers into an audit log for threshold self-calibration.
Owner:BEIJING FISKU SUPPLY CHAIN MANAGEMENT CO LTD

Image generation method for neural network model test

The invention relates to the field of model testing, and discloses an image generation method for a neural network model test, which comprises the following steps: acquiring a prediction result and confidence distribution of a target neural network on a standard test image set, identifying weaknesses of a model in texture, structure and illumination in combination with a sensitivity analysis algorithm, and obtaining a prediction result of the target neural network; constructing a model error-prone feature set; based on the error-prone feature set, performing semantic-preserving directional disturbance on the original image through a feature parameter injection method; evaluating the structural consistency between the generated image and the original image by using a perception similarity measurement and feature projection algorithm; a self-adaptive strategy is adopted to optimize a generation rule, a model feedback loop mechanism is introduced, and a disturbance parameter range is dynamically adjusted according to the identification error condition of the model on a newly generated image; and automatically generating a model performance diagnosis analysis report in combination with the model output change curve and the weakness dimension labeling result. The method has the advantage of improving the accuracy of model testing.
Owner:SHANGHAI YINYU DIGITAL TECH GRP CO LTD

Systems and methods for determining a location of a gross target volume of a patient

Provided herein are systems for determining a location of a gross target volume of a patient. In some examples, systems can include one or more processors that are configured to obtain image data associated with a plurality of images of a lesion of a patient. For each image, the one or more processors can be configured to backproject points representing the lesion into the 3D space to determine a plurality of distribution confidence values for a subset of voxels within the three-dimensional space. The one or more processors can be configured to determine a three-dimensional confidence distribution based on confidence values from the plurality of distribution confidence values corresponding to each voxel of the 3D space and determine a position of the lesion within the 3D space based on the 3D confidence distribution.
Owner:SIEMENS HEALTHINEERS INTERNATIONAL AG

Semi-supervised medical image segmentation method based on high-value region adaptive learning

The invention discloses a semi-supervised medical image segmentation method based on high-value region adaptive learning, and the method comprises the steps: constructing a semi-supervised medical image segmentation model based on high-value region adaptive learning, and the model comprises two parallel networks which have the same structure and independently update parameters; the two parallel networks output an original prediction result and a disturbed prediction result, and different value regions, namely a reliable and unstable region and an unreliable and unstable region, are screened out based on confidence distribution of the disturbed prediction result; training a reliable and unstable area through a confidence-guided cross prototype consistency learning module, and training an unreliable and unstable area through a dynamic teacher competition teaching module; and training the model by using the training set, selecting an optimal weight based on the verification set, inputting the test set into the model with the optimal weight, and outputting a result. Through the differential learning strategy, the method is accurately focused on the high-value region of the medical image, and the segmentation effect of the model in the key region of the image is remarkably improved.
Owner:SHAANXI UNIV OF SCI & TECH

Underwater robot sensor information fusion method and system based on evidence theory

The invention provides an underwater robot sensor information fusion method and system based on an evidence theory, and the method comprises the steps: collecting and preprocessing the measurement data of each sensor of an underwater robot under a single control algorithm, and obtaining data without obvious conflicts; based on the simplified JDL data fusion model and a Pearson coefficient, evaluating the data without obvious conflicts, and based on an evaluation result, performing initial confidence distribution on the data without obvious conflicts; checking the conflict degree between the initial confidence distribution based on the conflict coefficient, and combining and outputting the confidence of which the conflict degree meets a preset threshold based on a D-S evidence theory combination rule; and performing approximate normalization processing on the combined output based on the Bayesian theory to obtain the probability of accuracy of measurement data of each sensor, and completing underwater robot sensor information fusion based on the evidence theory. According to the method, the simplified JDL model and the Pearson coefficient are introduced to perfect the fusion mechanism, and the accuracy and rationality of fused data are improved.
Owner:GUANGDONG OCEAN UNIVERSITY

Power distribution system intelligent management method based on big data

The invention discloses a power distribution system intelligent management method based on big data, relates to the technical field of big data, and improves the accuracy and timeliness of power grid fault diagnosis. According to the method, a plurality of core point locations are set for historical operation data collected in different operation and maintenance periods and corresponding to the same operation and maintenance sub-period, the confidence degree of each core point location is obtained, and then standard operation data intervals of various types of operation data are obtained according to confidence degree distribution; the method comprises the following steps: intercepting abnormal historical data fragments under various fault item name combinations from historical operation data through a standard operation data interval, establishing an abnormal data vector combination according to the abnormal historical data fragments under different fault item name combinations, and judging whether the real-time operation data is abnormal or not through the standard operation data interval. And matching an abnormal data vector combination according to the type and the quantity of the real-time operation data judged to be abnormal, and further judging a fault item currently existing in the power distribution sub-area.
Owner:XIAMEN MINGHAN ELECTRIC

Method and apparatus for calibrating a language model

PCT designated stage expiredWO2025059913A9AlgorithmTheoretical computer science
The present disclosure provides a computer implemented method for calibrating a language model (LM), comprising: receiving, by a first LM, a set of language questions and generating a first set of predictive answer confidence distributions corresponding respectively to the set of language questions; receiving, by a second LM, the set of language questions and generating a second set of predictive answer confidence distributions corresponding respectively to the set of language questions, wherein the generating the second set of predictive answer confidence distributions comprising calibrating the second set of predictive answer confidence distributions based on a temperature scaling coefficient; determining a distance between the first set of predictive answer confidence distributions and the second set of predictive answer confidence distributions; and updating the temperature scaling coefficient based on the distance.
Owner:ROBERT BOSCH GMBH +1

Confidence evaluation method, device and equipment of large language model and storage medium

The invention discloses a confidence evaluation method and device for a large language model, equipment and a storage medium, and the method comprises the steps: obtaining a cue word set which comprises a plurality of cue words; sequentially inputting a plurality of cue words in the cue word set into the to-be-evaluated large language model to obtain a target sequence generated by the to-be-evaluated large language model based on each cue word, and obtaining the confidence coefficient of each target sequence according to the confidence coefficient of each lexical element in each target sequence; and obtaining a generation number of the target sequences, determining a confidence degree distribution result of the to-be-evaluated large language model according to the confidence degrees of all the target sequences when the generation number reaches a preset number, and generating an evaluation result of the to-be-evaluated large language model according to the confidence degree distribution result. Therefore, the problems that it is difficult for a user to effectively monitor the open source large language service model and it is difficult to provide evidence that the service model has degradation behaviors are solved, and the service quality of the open source large language model can be monitored on the user side.
Owner:TSINGHUA UNIVERSITY

Uncertain knowledge graph reasoning method based on semi-supervised confidence distribution learning

The application discloses an uncertainty knowledge graph reasoning method based on semi-supervised confidence distribution learning, comprising the following steps: converting the triple confidence in the uncertainty knowledge graph training data into a confidence distribution; simultaneously learning the embedding of the uncertainty knowledge graph on the labeled data and the pseudo-labeled data generated by a pseudo-labeled data generator by using a relation learner based on confidence distribution learning; generating high-quality pseudo-confidence distribution labels for unlabeled data by using the pseudo-labeled data generator; iteratively training the relation learner based on confidence distribution learning and the pseudo-labeled data generator by using meta-self-training until the two converge; and inputting the data to be completed into the trained relation learner based on confidence distribution learning to perform reasoning, thereby achieving the completion of the uncertainty knowledge graph. The application can capture the supervision information of a small number of confidence or unseen confidence in the labeled data and is suitable for the scene where the triple confidence distribution of the uncertainty knowledge graph is unbalanced.
Owner:SOUTHEAST UNIV

Systems and methods for determining a location of a general target volume of a patient

The present invention relates to systems and methods for determining a location of a gross target volume of a patient. In some examples, a system can include one or more processors configured to obtain image data associated with a plurality of images of a lesion of a patient. For each image, the one or more processors can be configured to backproject points representative of the lesion into a 3D space to determine a plurality of distribution confidence values for a subset of voxels within the three-dimensional space. The one or more processors can be configured to determine a three-dimensional confidence distribution based on the confidence values from the plurality of distribution confidence values corresponding to each voxel of the 3D space and determine a location of the lesion in the 3D space based on the three-dimensional confidence distribution.
Owner:SIEMENS HEALTHINEERS AG

Secondary equipment state evaluation method based on evidence theory

The invention discloses a secondary equipment state evaluation method based on an evidence theory. According to the method, the application advantages of the evidence theory in comprehensive evaluation of secondary equipment are considered and analyzed. Firstly, equipment health states are divided; and secondly, determining an information fusion model according to the index characteristics of each level, fusing evidences by adopting an ER rule and an ER algorithm for different index levels, and finally obtaining a comprehensive evaluation model of the equipment state. Thirdly, determining a model input quantity, acquiring evidence weight by adopting an improved G1 method, acquiring confidence distribution of evidence by adopting a model based on a ridge-shaped membership function, and acquiring evidence reliability by adopting an evidence distance; and finally, determining the final state of the equipment according to the degree of support of the equipment to each state obtained by the evaluation model and the maximum membership principle. The method is suitable for comprehensively evaluating the secondary equipment, and the health state of the secondary equipment can be more accurately evaluated, so that the method can be used for guiding the analysis design, operation maintenance and other work of the secondary equipment.
Owner:NORTH CHINA ELECTRIC POWER UNIV

Scene data classification method, scene data application method, electronic equipment, vehicle and medium

The invention relates to a scene data classification method, a scene data application method, electronic equipment, a vehicle and a medium. The classification method comprises the following steps: executing a first classification task on scene data to obtain first confidence distribution data; a first confidence coefficient and a first category corresponding to the first confidence coefficient are obtained from the first confidence coefficient distribution data, and the first confidence coefficient is the highest confidence coefficient in the first confidence coefficient distribution data; a second confidence coefficient is obtained from the first confidence coefficient distribution data, and the second confidence coefficient is a second highest confidence coefficient in the first confidence coefficient distribution data; obtaining a first certainty index value representing the certainty degree of the first confidence coefficient according to the first confidence coefficient and the second confidence coefficient; and under the condition that the first deterministic index value is greater than or equal to a preset threshold value, determining the first category as a classification result of the scene data.
Owner:CORECHENG (BEIJING) TECHNOLOGY CO LTD

An information fusion method for slope stability estimation

The present invention discloses an information fusion method for slope stability estimation. The present invention monitors the changes in groundwater level and surface crack width based on liquid level meters and crack meters, combines them into characteristic variables, uses a clustering algorithm to obtain a reference value set of the characteristic variables, and constructs a corresponding slope stability grade reference confidence distribution. After obtaining the monitoring sample online, the reference confidence vector is fused according to the activation degree to obtain the fused stability confidence distribution, and finally the stability of the slope is determined according to the confidence maximization principle. The present invention constructs the influencing factors affecting the stability of the slope into a multidimensional characteristic variable, and the characteristic variable will match all reference value sets to obtain a reference confidence distribution. The reference confidence distribution is corrected and recursively fused to achieve slope stability estimation. This method improves the limitations of traditional assessment and alarm methods that rely on single monitoring data and can only match local reference values and cannot fully reflect the movement state of the slope.
Owner:HANGZHOU DIANZI UNIV

Source-domain-free transfer learning target detection method based on double-adapter pseudo tag generation

The invention discloses a source-domain-free transfer learning target detection method based on double-adapter pseudo tag generation, and the method comprises the steps: firstly initializing a double-teacher model and a student model, and carrying out the data enhancement of a target domain image; introducing a confidence adapter, dynamically adjusting a confidence screening range through statistical analysis according to confidence distribution characteristics of each category, and generating candidate pseudo tags; introducing a category imbalance adapter, and fusing a confidence historical threshold and a current threshold according to the frequency of each category to obtain a current final confidence threshold; and finally, supervising student model training by using the screened pseudo labels, and updating double-teacher model parameters through index moving average to realize passive domain migration of the model. According to the source-domain-free target detection method based on double-adapter pseudo tag generation, the problems that pseudo tag confidence distribution is inconsistent and category samples are unbalanced are effectively relieved, and pseudo tag quality and cross-domain detection performance are remarkably improved.
Owner:HUNAN UNIV

Network message text element extraction method based on large language model integration, electronic device and computer-readable storage medium

The invention relates to a network message text element extraction method, an electronic device and a computer-readable storage medium based on large language model integration, which belong to the field of text information element extraction. In order to solve the problem of integrating at element granularity, timely filtering and identifying erroneous elements, and improving the accuracy of generated elements, the technical key points are to use the confidence of any candidate element on any base model and the confidence of the termination token as the confidence of the proposition of the identification framework of the evidence reasoning rule, and generate evidence of any base model according to the confidence of the proposition of any base model in the identification framework; calculate the weighted confidence distribution of the evidence of any base model according to the evidence of any base model and the weight of any base model; according to the weighted confidence distribution of the evidence of any base model, fuse the evidence of all base models to obtain fused evidence, and use the element with the largest probability in the fused evidence as the integration result, so as to improve the accuracy of the extraction result.
Owner:DALIAN UNIV OF TECH

Face image classification method and device, electronic equipment, medium and program product

The invention provides a face image classification method and device, electronic equipment, a medium and a program product, and can be applied to the technical field of artificial intelligence and the field of financial science and technology. The method comprises the steps that a to-be-classified face image is acquired, multi-layer feature information of the face image is extracted based on a pre-trained depth feature extraction model, and the multi-layer feature information comprises feature mapping results output by different layers of the depth feature extraction model; performing fusion processing on the multi-layer feature information to obtain fused feature representation; the fusion feature representation is input into a classification model, multiple face categories and corresponding confidence distribution are output, a dynamic classification threshold value is calculated according to the confidence distribution, and the dynamic classification threshold value is used for adjusting classification boundaries of the multiple face categories; and obtaining the category of the face image based on the plurality of face categories and the dynamic classification threshold.
Owner:INDUSTRIAL AND COMMERCIAL BANK OF CHINA

Trust Evidence Reasoning Method, Apparatus and Device for Multi-Source Data Fusion

ActiveCN120012944BMachine learningInference methodsEvidence reasoningSupport matrix
The present application relates to a credible evidence reasoning method, apparatus and equipment for multi-source data fusion, which constructs a multi-source decision information system. In this multi-source decision information system, a support matrix is designed to quantify the correlation between multi-source data attributes, and the concepts of similarity and distance are combined to evaluate the relationship. First, multi-source data of different scales are converted into confidence distributions and given probabilistic meanings, so that the generated evidence can be applied to multi-scale data fusion. Then, a credible evidence reasoning rule is constructed, in which the credibility of the evidence is determined by its weight and reliability. By integrating multi-source data samples of different scales, the confidence distribution of the results is derived. This solves the problem of difficulty in determining the confidence distribution of evidence.
Owner:NAT UNIV OF DEFENSE TECH

A network performance evaluation method and device based on dynamic evidence reasoning rules

ActiveCN119945917BInference methodsTransmissionEvaluation resultEvidence reasoning
The application discloses a network performance evaluation method based on dynamic evidence reasoning rules, which comprises the following steps: dynamically adjusting the reference value of a test index based on the change of a network state; using the adjusted reference value to convert monitoring data into a confidence distribution form, converting the test index into a corresponding evidence form according to the mapping relationship of IF-Then; fusing the evidences corresponding to multiple indexes based on the evidence reasoning rules to obtain an evaluation result; and optimizing the model parameters by using an optimization algorithm through setting the constraint conditions and optimization targets of parameters. The application further discloses a network performance evaluation device based on dynamic evidence reasoning rules. The network performance evaluation method based on dynamic evidence reasoning rules can more accurately reflect the actual performance of a network in different scenarios according to the change of the network state, and meets the evaluation requirements of different network environments and business types.
Owner:NAT UNIV OF DEFENSE TECH

General model stealing defense method and system based on model decision boundary

The invention provides a universal model stealing defense method and system based on a model decision boundary, and the method comprises the steps: a training stage: obtaining a user prediction sample, taking the sample and a confidence coefficient distribution vector of a cloud platform model for the prediction sample as input, and outputting an original disturbance sample; adding the original disturbance sample to a user prediction sample as an input to obtain an iterative disturbance sample; in the reasoning stage, when the prediction label obtained by the original confidence distribution vector is the same as the prediction label obtained by the hidden confidence distribution, iteration is continued. When not, iteration is stopped, and the distance from the prediction sample to the decision boundary is determined according to the number of iterations; determining a malicious sample according to the obtained distance and a given comparison result; and returning a malicious sample disturbance confidence distribution vector, and limiting user access. According to the method, under the condition that the output label of the target model is not changed, the output confidence distribution vector of the target model is changed to the greatest extent, so that the availability of the stolen model is reduced.
Owner:SOUTH CENTRAL UNIVERSITY FOR NATIONALITIES

Feature classification method and system based on alimentary canal internal image

The invention discloses a feature classification method and system based on an alimentary canal internal image, and relates to the technical field of image recognition, and the method comprises the steps: collecting an alimentary canal image in real time through an endoscope, and carrying out the preprocessing; based on a position estimation network formed by a convolutional neural network and a time sequence encoder, the image sequence is mapped to a standardized alimentary canal template, and endoscope position evaluation is achieved; the attention weight of each sample is calculated according to the time sequence similarity of the similar positions, and individual adaptive optimization of the image recognition model is completed; outputting an identification result, severity and confidence distribution of a focus type through an image identification model, constructing a graph structure model taking the focus type as a node, performing information propagation in a graph structure, and deducing focus probability distribution of an undetected area; and an inference result is fed back to the attention weight of the classification output layer, so that dynamic self-learning and global optimization of the model are realized. According to the method, the accuracy and stability of alimentary canal image recognition are remarkably improved.
Owner:GUIZHOU MEDICAL UNIV