63 results about "Misclassification error" patented technology

The invention discloses a method and a system for monitoring application logs. The method comprises the steps of capturing wrong information by an application server when exception occurs in a running application program, preprocessing the wrong information to generate an application blog, outputting the application blog to a monitoring server through a Scribe interface, receiving the application blog by the monitoring server through the Scribe interface to conduct sorted storing, reading the stored application log and obtaining the wrong information according to preset classification error types, then judging whether the obtained wrong information satisfies preset alarm strategies or not, and sending the obtained wrong information to a preset target user. By the application of the method and the system for monitoring the application logs, monitoring efficiency of the application blog can be improved, and real-time quality of application blog processing is promoted.

Disclosed are a variety of methods and computer systems for use in the analysis of gene and protein expression data. Also disclosed are methods for the definition of the cellular state of cells and tissues from multidimensional physiological data such as those obtained from gene expression measurements with DNA microarrays. A variety of classification methods can be applied to expression data to achieve this goal. Demonstrated is the application of several statistical tools including Wilks' lambda ratio of within-group to total variance, Fisher Discriminant Analysis, and the misclassification error rate to the identification of discriminating genes and the overall classification of expression data. Examples from several different cases demonstrate the ability of the method to produce well-separated groups in the projection space representing distinct physiological states. The method can be augmented and is useful in disease diagnosis, drug screening and bioprocessing applications.

The invention discloses a 3D convolutional neural network sign language identification method integrated with multi-modal data. The 3D convolutional neural network sign language identification methodintegrated with multi-modal data includes the steps: constructing a deep neural network, respectively performing characteristic extraction on an infrared image and a contour image of a gesture from the spatial dimension and the time dimension of a video, and integrating two network output based on different data formats to perform final classification of sign language. The 3D convolutional neuralnetwork sign language identification method integrated with multi-modal data can accurately extract the limb movement track information in two different data formats, can effectively reduce the computing complexity of a model, uses a deep learning strategy to integrate the classification results of two networks, and effectively solves the problem that single classifier encounters error classification because of data loss, so as to enable the model to have relatively higher robustness for illumination and background noise of different scenes.

The invention provides a robust gun detection method. According to the method, effective pretreatment and target significance detection are performed, and the candidate areas of a target detection object are searched through color segmentation and screening of specific dimension characteristics. According to the method, a gun classifier is designed based on the gun and non-gun classification error risk, then multiple multilayer deep learning neural networks are cascaded, and finally discrimination is performed by statistics of the gun possibility through the candidate areas so that finally a complete gun detector is formed. The gun detection method is great in anti-interference performance, high in real-time performance and high in identification precision.

The invention provides an increment study classification method under limited storage resources, which belongs to the technical field of pattern recognition. The method adopts a minimum distance classifier and comprises the following concrete steps: firstly, pre-classifying new samples, adding the new samples which are correctly pre-classified to the corresponding subset of the classifier, adding the new samples which are wrongly pre-classified to the set of wrong samples, and carrying out K-mean clustering on the samples in the set of wrong samples; then respectively selecting representative samples for each subset in the classifier and the set of wrong samples, adding the subsets in the set of wrong samples to the classifier after selection, and updating the classifier; and finally, adopting the updated classifier to classify the new samples. In the invention, by selecting the representative samples, not only the studied knowledge is saved, but also the new knowledge is acquired, and higher sample recognizing accuracy is achieved on the basis of reducing the storage cost and calculating the cost.

The invention belongs to the field of image recognition, particularly relates to an adversarial sample generation method, system and device for an outlier removal method, and aims to solve the problems that an adversarial sample adopted by existing classification model training based on deep learning cannot make an image classification error under an outlier removal method; and therefore, the trained classification model is poor in robustness and low in accuracy. The method comprises the following steps: acquiring a training data set with category labels, inputting three-dimensional point cloud data into a classification model, calculating classification loss, respectively calculating the gradient of the classification loss relative to the three-dimensional point cloud data and the gradient of the classification loss relative to outlier-removed three-dimensional point cloud data, and fusing the two gradients by multiplying a scaling factor to generate fusion disturbance, and applying the fusion disturbance to the three-dimensional point cloud data for repeated iteration to generate an adversarial sample. The generated adversarial samples can still cause image classification errorsunder the condition that outliers are removed, and the robustness and classification accuracy of the trained model are improved.

Infliximab (IFX) is an effective treatment for Crohn's disease (CD) and ulcerative colitis (UC) not responding to standard therapy. Thirty percent to forty percent of patients however do not improve and the response is often incomplete. We identified mucosal gene signatures predictive of response to EFX using high-density oligonucleotide arrays. Eight UC patients and twelve CD patients showed healing. In UC, only one probe set was differentially expressed in responders compared with non-responders, i.e., IL-13R(alpha)2. At PAM analysis, two probe sets, representing IL-13Ralpha2 and IL-I 1, separated IBD responders from non-responders with an overall misclassification error rate of 0.046 (2 / 43), with 100% sensitivity and 91.3% specificity. The IL-13R(alpha)2 probe set was a top-ranked probe set in all our analyses using both LIMMA and PAM strategies. Our gene array studies of mucosal biopsies identified IL-13R(alpha)2 in IBD as a predictor of response or non-response to IFX.

The invention discloses a multi-scale fuzzy measure and semi-supervised learning based SAR (Synthetic Aperture Radar) image identification method and solves the problem that the SAR image identification accuracy in the prior art is low. The multi-scale fuzzy measure and semi-supervised learning based SAR image identification method comprises the following steps of establishing an image library by segmenting an original SAR image and selecting image blocks with single targets; extracting characteristic vectors of the image blocks in the image library; classifying the selected image blocks into a plurality of categories, enabling corresponding characteristic vectors to be served as training samples, training a semi-supervised classifier and classifying the image library through the classifier; obtaining categories of inquire image blocks input by a user through a trained classifier; obtaining a category set of the image blocks through a confusion matrix; calculating a multi-scale area fuzzy similarity between the inquire image blocks and the image blocks belong to the set and returning the number of user required image blocks according to a sequence from large to small. The multi-scale fuzzy measure and semi-supervised learning based SAR image identification method can correct the classification error, is high in information identification accuracy and can be applied to simultaneous explain of a plurality of SAR images.

The invention discloses a high-spectral unsupervised classification method for constructing a generic dictionary based on confidence degrees. The method comprises the steps of firstly, constructing a two-dimensional spectral-pixel matrix; performing row and column standardization processing; performing feature extraction and selection to obtain dimension reduction features of pixels; performing coarse classification and confidence degree assessment, namely, classifying the pixels by utilizing the dimension reduction features, calculating Euclidean distances between the spectral pixels and a coarse classification category center to serve as the confidence degrees, and obtaining high-confidence-degree classified samples and low-confidence-degree classified samples; and finally, performing secondary classification based on kernel sparse representation, namely, forming the generic dictionary by the high-confidence-degree classified samples, performing the kernel sparse representation on the low-confidence-degree classified samples, and determining classification tags of the low-confidence-degree spectral pixels. According to the method, the problems of low classification sub-space description precision and excessively high computing complexity due to construction of the dictionary by directly utilizing all spectral data are solved; the dictionary sub-space identification performance is improved; and the misclassification error rate is reduced.

An online hand-written chemical symbol identification method based on the Hidden Markov model solves the problem of the online identification of chemical symbol written by any one writer on any one device. The method constructs a processing framework of identification of the online written chemical symbol, and employs a hierachical processing and step-by-step optimization strategy. The method based on a support vector machine selects grid features and peripheral contour features to distinguish organic ring symbols and non-ring symbols, and the classification error rate is controlled under 0.2%. The method based on the Hidden Markov model to identify concrete symbols, and the accuracy is more than 90%. In order to improve the precision design, a set of preprocessing flow is designed, and the post-processing measures such as candidate result reliability, chemical symbol adjacent matrixes and the atom element conservation detection are employed. The method has the universal meaning, systematicness and completeness through the experimental proof of the data source such as the input of a Tablet PC, a digital panel and a mouse simulation pen, and can be used for the online hand-written chemical symbol identification field.

The invention discloses a deep learning satellite data cloud detection method supported by hyperspectral data. The method comprises the following steps: selecting enough cloud and clear sky pixels toconstruct a hyperspectral data sample library, and performing analog computation on the hyperspectral pixel sample library according to parameters such as a spectral response function and a waveband width of a to-be-detected sensor to obtain a cloud and clear sky surface pixel library of the to-be-detected sensor; based on a Keras deep learning framework, designing a deep BP neural network for cloud detection, inputting multispectral sample data obtained through simulation into the network for training and learning, and obtaining a multispectral sensor cloud detection rule based on spectral characteristics; based on a Markov random field model, optimizing a cloud detection result by utilizing an iterative condition mode algorithm, and removing a part of misclassification and misclassification errors of cloud detection. According to the method, various sensor data are selected and compared with a cloud coverage result of artificial visual interpretation for analysis, and the result shows that the algorithm achieves a better cloud detection effect and can meet the requirements of data application for cloud detection.

The invention provides a concept drift detection method based on a classification error rate and consistency prediction, and belongs to the technical field of computer machine learning and informationsecurity. According to the method, mutation type concept drift is detected by calculating a change of the classification error rate of a model, and then the progressive concept drift is detected by calculating a consistency degree of the samples with wrong classification and the samples with correct classification so that the mutation type concept drift and the progressive concept drift can be detected in time, and relatively low calculation overhead is kept. According to the method, detection of mutation type concept drift and progressive concept drift is achieved at low calculation cost, and a model degradation phenomenon is recognized in time. The method is mainly used for concept drift detection, can effectively act on early judgment of a degradation phenomenon of a machine learning classification model, and can be used as a performance monitoring method in various application fields such as automatic analysis and decision in a big data environment.

The invention relates to an optical fiber sensing vibration signal mode recognition method based on an AdaBoost-ESN algorithm. The method comprises the following steps: collecting vibration signals toconstruct a data set; constructing a pattern recognition network of an AdaBoost-ESN algorithm; classifying the vibration signals of a plurality of known events; the method comprises the following steps: firstly, constructing an ESN basic classifier by using an echo state network; a basic classification result is obtained by training and identifying samples, then an AdaBoost iterative framework isadopted to calculate classification error samples of an ESN basic classifier, the weight of the classification error samples is improved, a next ESN basic classifier is constructed, and after multiple iterations, a final strong classifier is output in a centralized mode.

The present invention relates to the technical field of manipulators, in particular to a mechanical separation device, which includes a supporting connecting plate, and a staggering mechanism is arranged on the upper end of the supporting connecting plate, and the staggering mechanism includes a staggering base, and the staggering base is movably connected There are staggered blocks, the staggered blocks are provided with a discharge port, and the staggered base is provided with a dislocation hole cooperating with the discharge port, and the dislocation hole is connected with a good product discharge channel and a defective product discharge One end of the staggered base is connected with a staggered cylinder; the invention is simple in structure and saves space. When in use, the product is put into the discharge port through the mechanism, and the staggered cylinder drives the staggered blocks in the staggered base according to the detection results. Move, make the product fall into the feeding channel, realize the separation and feeding of good products and bad products, so as to achieve good synchronization, fast beat, less power, save a lot of air pipes and signal lines, simple and convenient control, beautiful wiring, and low cost Purpose.

The invention belongs to the technical field of Internet, and discloses an e-commerce commodity classification method and system based on a hierarchical combination model, and the method comprises thesteps: building a multi-layer classification lexicon or a hierarchical classification lexicon based on the hierarchical combination model; and matching the commodity titles subjected to word segmentation processing with a multi-layer classification word bank or a hierarchical classification word bank step by step from the uppermost level to the lower level, and finally determining the classification of the commodities by using paths or hierarchical classification tags according to a matching result. According to the classification method and system, on one hand, classification confusion of different categories of commodities with the same name is effectively avoided, and on the other hand, the problem of classification errors caused by noun stacking in titles is successfully solved.

The invention relates to the technical field of reservoir lithology identification, in particular to a method for identifying stratum lithology parameters based on multi-kernel ensemble learning. Themethod comprises the steps of dividing different sample sets according to logging parameter characteristics; dividing a training sample set and a test sample set; establishing a strong classifier forlithology parameter characteristics, judging test samples in the test sample set, and obtaining lithology parameters by adopting an average method; establishing a strong classifier for the lithology parameter characteristics according to a prediction result and reconstructed sample data; forming a strong classifier by utilizing the strong classifier; judging the samples, and determining a final stratum lithology category by adopting a voting mode; adopting an absolute majority voting method, if a certain lithology is marked to obtain a half vote, prediciting the lithology, and otherwise, refusing prediction. Characteristics of a multi-kernel ensemble learning algorithm are applied, a plurality of classifiers are combined, the classification error rate is minimized, the logging data utilization rate is improved, and the judgment accuracy is high.

The invention discloses an instance segmentation method for correcting classification errors by using a classification attention module. The method comprises the steps of obtaining a plurality of feature maps of a to-be-processed image based on a backbone neural network of a preset instance segmentation model; performing convolution processing on the feature map based on a classification module ofa preset instance segmentation model to obtain a semantic category of the to-be-processed image; performing convolution processing on the feature map based on a classification attention module of a preset instance segmentation model to obtain a pixel category of the to-be-processed image; determining a foreground class channel of the to-be-processed image based on the pixel foreground class channel and the semantic class; and performing convolution processing on the mask convolution kernel parameter of the foreground class channel of the to-be-processed image and the mask feature map so as toobtain the prediction mask of the foreground class channel of the to-be-processed image, instance segmentation is performed on the image according to the prediction mask, and therefore, misclassification instances can be corrected, and the accuracy of image instance segmentation is further improved.

The invention relates to the field of model training, in particular to a training data enhancement method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a training data set; inputting the training data set into a classification model for pre-training, and obtaining an original image and classification data which are wrongly classified; inputting the original image into a preset residual network to obtain features of the original image; calculating mutual information of the features according to the classification data, and screening false features of which the influence degree is greater than a preset value in the features according to the mutual information; performing data enhancement on the original image according to the false features to obtain a target image; and generating an enhanced training data set according to the target image, and retraining the classification model based on the enhanced training data set. According to the method and the device, errors of the model can be traced, features related to the prediction result can be searched, the original data set is enhanced, and the robustness of the model is improved.

A system for evaluating fault criticality using machine learning includes a first machine learning module that is trained on a subset of a circuit and used for evaluating whether a node in a netlist of the entire circuit is a critical node, and a second machine learning module specialized to minimize classification errors in nodes predicted as benign. A generative adversarial network can be used to generate synthetic test escape data to supplement data used to train the second machine learning module.

Techniques are disclosed relating to improving machine learning classification using both labeled and unlabeled data, including electronic transactions. A computing system may train a machine learning module using a first set of transactions (of any classifiable data) with label information that indicates designated classifications for those transactions and a second set of transactions without label information. This can allow for improved classification error rates, particularly when additional labeled data may not be present (e.g., if a transaction was disallowed, it may not be later labeled as fraudulent or not). The training process may include generating first error data based on classification results for the first set of transactions, generating second error data based on reconstruction results for both the first and second sets of transactions, and updating the machine learning module based on the first and second error data.

The invention discloses a method for classifying electrocardiosignals by using an optimized AdaBoost weighting mode and a weak classifier, and belongs to the technical field of data source classification. The method comprises the steps: conducting feature extraction on electrocardiogram data to obtaining an electrocardiogram signal training sample data set; according to the initial weight of the electrocardiosignal sample, sampling and selecting an electrocardiosignal training data set to obtain the electrocardiosignal sample; using the weak classifier for classifying electrocardiosignal samples, wherein the weighting mode of the samples which are wrongly classified is determined by the classification error rate of the weak classifier and the probability that the samples are wrongly classified. According to the method, the misclassification probability of misclassification samples is fully considered, and the weak classifier is screened by setting a proper threshold value, so that thestrong classifier with higher classification precision is obtained.

The invention relates to a cross-domain sentiment analysis method based on a comparison and alignment network, and belongs to the technical field of fine-grained sentiment analysis in natural language processing. According to the method, a scene, which is not fully explored, of cross-domain sentiment classification is researched, that is, a target domain is a scene with few samples. In the scene, the invention provides a neural network model named as a contrast alignment network (CAN). According to the model, two instances are randomly extracted from an original domain and a target domain, and then the target domain and the original domain are trained according to the instances of the combined target domain and the original domain. The first objective is to minimize classification errors on the original domain. The second is a pair of contrast targets, where a measure of distance between a target domain instance and an original domain instance in one pair is minimized if they express the same emotion, otherwise the measure is maximized with a constant upper limit. According to the method, the problem that the target domain data resources in the cross-domain sentiment classification task are limited is solved, and the user experience is improved.

The invention relates to a data flow classification algorithm based on AAE-DWMIL-LearnNSE, which is a method for classifying data flows, belongs to the field of data mining and machine learning, and is characterized by comprising the following steps of: (1) updating the weight of each classifier; (2) carrying out weighted processing on the base classifier; (3) weighting and synthesizing an integrated classifier; (4) calculating a classification error rate on the new data set; (5) creating a comprehensive prediction model; and (6) determining and classifying a data flow classification function.The problem that the classification precision is greatly influenced due to lack of selective selection of the training samples is solved, the link of storing old data samples is omitted, most of storage space is saved for new data, meanwhile, an old classification model is fully utilized, and high classification accuracy is obtained. The problem of similarity between labeled training samples in the measurement source field and unlabeled test samples in the target field is effectively solved.

The invention discloses a tunnel traffic accident duration prediction method based on PCA and Adaboost, and the method comprises the following steps: importing historical traffic accident data: carrying out the preprocessing of the data, and dividing the data into a short grade, a medium grade, a long grade and an extra-long grade according to the duration of an accident; carrying out the missing value checking and processing on input variables in the prediction model; and finally, carrying out thermal coding processing on the classification variables. Herein, the PCA method is used for decentralizing original input variables and calculating a covariance matrix of the original input variables, and feature values and feature vectors of the original input variables are calculated on the basis, and a plurality of feature values and corresponding feature vectors are sequentially determined from small to large. Firstly, traffic accident duration is classified based on a weak classifier, and a basic classification result is obtained through sample training; and then, an Adaboost iteration framework is adopted to calculate classification error samples of the weak classifier, the weight of the classification error samples is improved, a next weak classifier is constructed on this basis, and a final strong classifier is obtained after multiple iterations.

The invention provides a data fusion method and device, electronic equipment and a computer readable storage medium, and the method comprises the steps: classifying to-be-fused data from different sources, and obtaining a plurality of subclass data; searching similar data of each sub-category data in a pre-established standard database, and determining the similarity between the similar data and the sub-category data; and fusing the sub-category data with a standard database according to the similarity. According to the method, the classification error rate can be reduced, invalid data in thefusion database is reduced, and the quality of database data is improved.

According to the data stream classification method and device, the concept drift state is accurately recognized according to the classification error rate of the integrated classifier at the current moment and the concept drift detection threshold value, the data window at the next moment is flexibly set according to the concept drift state, and if the data window is rapidly increased when concept drift does not occur, the data window can be rapidly increased. The data stream classification efficiency is improved; when it is not determined whether concept drift occurs or not, the increasing amplitude of a data window is reduced, whether concept drift occurs or not is further determined, and the influence of noise on classification accuracy is reduced; when concept drift occurs, a data window is quickly reduced, the sensitivity to sudden change type concept drift is improved, and the integrated classifier is quickly adjusted, so that the accuracy of data stream classification by the integrated classifier is integrally improved.

The invention discloses a garbage classification processing method, related equipment and a readable storage medium. The garbage classification processing method determines information of a commodityby identifying an image of the commodity to be subjected to garbage classification processing, and then analyzes the information, so as to determine an object generated after the commodity is used anda garbage type corresponding to the object. Based on the determined object generated after the commodity is used and the garbage type corresponding to the object, a user can determine the object generated after the commodity is used and the garbage type corresponding to the object, and garbage classification is carried out based on the object generated after the commodity is used and the garbageclassification error probability can be reduced.

The invention discloses a forgetting verification method based on semantic data loss in federated learning, and the method employs specific performances on some data which are high in loss, are commonly wrong, and are provided with a certain semantic feature to mark forgetting users and verify forgetting conditions. The method comprises the following steps: screening out data with high loss and general classification errors in a local data set, and re-marking the data as a certain fixed category according to similarity and confidence distribution of semantic features of the data to obtain a marked data set; and uploading the marked model after the local model is finely adjusted on the data set and the original data set to a central server for aggregation. The forgetting user verifies the forgetting condition according to the loss of the global model on the marked data set by checking the global model of the next plurality of cycles. The method has the advantages of being lightweight, high in continuity, good in verification effect, small in time and space overhead and the like, can effectively identify whether forgetting occurs or not, and can be widely applied and deployed in various scenes needing forgetting verification.

The invention discloses an image classification method and device, which are used for solving the problem that relatively similar images cannot be accurately classified. The method comprises the stepsof inputting a preset classification image into a main model to obtain a first classification result; inputting the first classification result into the condition judgment model to obtain judgment parameters; when the judgment parameter is greater than or equal to a preset threshold, outputting the first classification result; and when the judgment parameter is smaller than a preset threshold, inputting the preset classification image into the error correction model to obtain a second classification result, and outputting the second classification result. In the embodiment of the invention, the condition judgment is introduced to correct the classification result of the main model, so that the probability of classification errors is reduced, and the accuracy of image classification is improved.

The invention belongs to the field of image recognition, and specifically relates to an adversarial sample generation method, system and device for outlier removal methods, aiming to solve the problem that the adversarial samples used in the existing deep learning-based classification model training cannot The method of removing outliers makes image classification errors, resulting in the problem of poor robustness and low accuracy of the trained classification model. The present invention includes: obtaining a training data set with a class label, inputting three-dimensional point cloud data into a classification model and calculating a classification loss, respectively calculating the gradient of the classification loss on the three-dimensional point cloud data and the three-dimensional point cloud data on removing outliers The gradient of the two gradients is multiplied by the scaling factor to generate fusion perturbation, and the fusion perturbation is applied to the three-dimensional point cloud data to iteratively generate adversarial samples. The adversarial samples generated by the invention can still cause image classification errors under the condition of removing outliers, which improves the robustness and classification accuracy of the trained model.