369 results about "Sample Weight" patented technology

The present invention is generally directed to various methods and systems for adaptive metrology sampling plans that may be employed to monitor various manufacturing processes. In one example, the method includes creating a plurality of metrology sampling rules, assigning each of the metrology sampling rules a sampling weight value, identifying at least one workpiece that satisfies at least one of the metrology sampling rules, assigning the sampling weight value for each of the satisfied metrology sampling rules with the identified workpieces that satisfy the rules, and indicating a metrology operation should be performed when a cumulative total of the sampling weight values is at least equal to a pre-established trigger value.

The invention relates to a robust ear detection method which improves an AdaBoost arithmetic and belongs to the technical field of image mode identifying. The invention is characterized by proposing an ear detection method with excellent performances under a complex underground. The invention proposes four anisomerous Haar-like corner characteristics which are used for describing the grayscale changes on the partial areas of the ears; a policy of subsection selection is adopted for selecting the best sorting threshold of the Haar-like characteristics, thus reducing the sample training time; the weight of a weak sorter is modified for reducing the mistaken detection rate of the sorter; the threshold HW is set and eliminated according to the distribution change of the sample weight in the training, thereby preventing an over-studying phenomenon from being generated and leading the miss-detection rate and the mistaken direction rate of the ear detection to be reduced; besides, the invention also provides a single-ear detection policy for leading both the defection efficiency and the detection effect to be improved. The excellent performances of the robust ear detection method are shown on a PC machine and a DSP.

The invention discloses an unbalanced data classification method based on unbalanced classification indexes and integrated learning, and mainly solves the problem of low classification accuracy of the minority class of the unbalanced data in the prior art. The method comprises steps as follows: (1), a training set and a testing set are selected; (2), training sample weight is initialized; (3), part of training samples is selected according to the training sample weight for training a weak classifier, and the well trained weak classifier is used for classifying all training samples; (4), the classification error rate of the weak classifier on the training set is calculated, is compared with a set threshold value and is optimized; (5), voting weight of the weak classifier is calculated according to the error rate, and the training sample weight is updated; (6), whether the training of the weak classifier reaches the maximum number of iterations is judged, if the training of the weak classifier reaches the maximum number of iterations, a strong classifier is calculated according to the weak classifier and the voting weight of the weak classifier, and otherwise, the operation returns to the step (3). The classification accuracy of the minority class is improved, and the method can be applied to classification of the unbalanced data.

The invention relates to a category weight combined integrated learning classifying method. The method comprises the steps of preprocessing original data; converting into data formats that can be processed by the classification method so as to obtain a training data set and a data set to be classified; initializing the training data set sample weight; re-iterating and training M base classifiers according to the training data set and the sample weight thereof; calculating the category weight; integrating all base classifies; classifying the data set to be classified through a determining classifier according to the category weight; storing the classifying result into a file to obtain classification predication reference. With the adoption of the method, the problem of unbalancing category training under multi-category and multi-classification condition of integrated learning can be solved, the excessive learning is inhibited well, and moreover, the model predication precision is improved, and reliable reference is provided for classification predication.

The invention discloses a pedestrian detection model training method based on an AdaBoost classifier. The pedestrian detection model training method comprises the steps of firstly, conducting real-time statistics on the sum of sample weight values in the AdaBoost training process, when degeneration is carried out to a certain extent, using a currently-trained weak classifier set for scanning a non-pedestrian image for a false detection window, using the false detection window as a difficult sample to be added in negative sample training sets, and decreasing a degeneration degree threshold value so as to reduce sample update efficiency; finally, removing a part of negative samples through random sampling, and reducing the number of the negative sample training sets so as to reduce the calculated amount of the training process. According to the pedestrian detection model training method based on the AdaBoost classifier, on the premise that a feature extraction method is not changed, the training effect of the classifier can be improved to the maximum extent, and the final detection precision is improved.

The invention relates to an unbalanced data sampling method in an improved C4.5 decision tree algorithm. The method comprises the steps as follows: firstly, initial weights of various samples are determined according to the number of various samples; the weights of the samples are modified through the training result of the improved C4.5 decision tree algorithm in each round; the information gain ratio and misclassified sample weights are taken into account by a division standard of the improved C4.5 algorithm; the final weights of the samples are obtained after T iterations; the samples in minority class boundary regions and majority class center regions are found out according to the sample weights; over-sampling is carried out on the samples in the minority class boundary regions by an SMOTE algorithm; and under-sampling is carried out on majority class samples by a weight sampling method, so that the samples in the center regions are relatively easily selected to improve the balance degree of different classes of data, and the recognition rates of the minority class and the overall data set are improved. According to the unbalanced data sampling method in the improved C4.5 decision tree algorithm, weight modification is carried out through the improved C4.5 decision tree algorithm; and over-sampling and under-sampling are specifically carried out according to the sample weights, so that the phenomena of classifier over-fitting, loss of useful information of the majority class and the like are effectively avoided.

The invention relates to a method and a device for weighting a living animal. The method for weighting the living animal comprises the following steps: sampling weight information of the living animal by use of an electronic scale according to a preset sampling frequency; acquiring weight information sampled within a preset time interval by use of the electronic scale; taking the weight information sampled within a preset time interval as a set of basic data for counting the weight under the condition that the electronic scale confirms that the stability of the weight information sampled within a preset time interval meets the demand on the preset stability; and calculating the weight of the living animal by use of the electronic scale according to preset sets of basic data. The technicalscheme of the invention can be used for ensuring the weighting error of the living animal to be no more than 0.1%. Compared with the weighting error, namely 1%, of the living animal in the prior art,the weighting error of the living animal is increased by an order of magnitude, so that the method for weighting the living animal can be used for efficiently improving the accuracy of weighting the living animal.

The invention discloses a federation modeling method based on privacy protection. The federation modeling method comprises the following steps: aligning local sample data of each modeling node to construct a training sample; initializing to-be-trained model parameters of each modeling node; creating an encryption key pair, and sending the public key to each modeling node; controlling each modelingnode to encrypt according to the public key, and interacting with an intermediate result for calculating an encryption gradient and encryption loss; receiving a joint encryption loss summarized and calculated by a specified modeling node; distributing the joint encryption sample weight summarized and calculated by the specified modeling node to other modeling nodes to calculate an encryption gradient; decrypting the encryption gradient calculated by each modeling node; and returning the decrypted gradient to each modeling node so as to update model parameters for training until the joint lossfunction is converged. The invention further provides a federation modeling device based on privacy protection and a computer readable storage medium. According to the invention, joint modeling can be carried out under the condition that the data of each modeling node is not leaked.

The invention discloses a transfer learning design method and system based on domain adaptation under a multi-example multi-label framework. According to the invention, multi-example multi-label learning and transfer learning are unified into one framework, source domain data samples and target domain data samples are effectively utilized for correlation statistics, and the source domain samples can be effectively used in the learning of target domain tasks; the characteristics of a source domain data sample set and a target domain data set in RKHS are utilized, a two-step domain adaptation process formed by sample weighting and a sample selection mechanism based on clustering is utilized, so that the learning of target tasks has enough training samples weighted and selected from the source domain set; and a miFV algorithm is utilized to convert multi-examples into a single example, and the calculation cost problem of domain adaptation is solved.

The invention relates to a method for extracting and calculating a capacitance parameter based on a random walk in an integrated circuit design, belonging to the technical field of integrated circuit computer aided designs, and comprising the following steps of: (1) setting the combination of the upper and lower dielectric constants of each interface in an integrated circuit and setting up a cubic transition area model, obtaining the relationship between the potential of an area surface grid and a central potential by a numerical way and taking the relationship as an initial transition probability intensity vector, and converting the initial transition probability intensity vector into a transition intensity vector and a corresponding weight numerical vector according to an importance collecting idea, and then processing and storing the transition intensity vector and the corresponding weight numerical vector into a database; and (2) correspondingly modifying a weight sampling way and a weight numerical value in a random walk algorithm by the data calculated in Step (1), and extracting and calculating the capacitance parameter in the integrated circuit. The random walk sampling weight numerical value generated by the method is unified, and the sampling probability tends to a location contributing great to the Gauss surface integral. The method for extracting and calculating the capacitance parameter has higher calculating efficiency and the design period for the integrated circuit is shortened.

The invention discloses a click and vision fusion based weak supervision bilinear deep learning method, which comprises the steps of 1, extracting click features of text composition of each image from a click dataset, and building new low-dimensional compact clock features in a combined text space through combining texts with similar semantics; 2, building a deep module with click and visual features being fused; 3, performing BP learning on network model parameters; 4, calculating the model prediction loss of each training sample, building a similarity matrix of the sample set, learning the sample reliability by using the sample loss and the similarity matrix at the same time, and weighting the samples by using the reliability; and 5, repeating the step 3 and the step 4, iteratively a neural network model and sample weights so as to train the whole network model until convergence. According to the method, click data and visual features are fused so as to construct a new bilinear convolution neural network framework which can be used for better performing recognition on a fine-grained image.

The invention discloses a method for forecasting flood based on a Boosting algorithm and a support vector machine, which comprises following steps: use the correlation coefficient method to determine the forecast factors; utilize kernel principal component analysis to process the forecast factors with dimension reduction; utilize the Boosting algorithm to select a sample and establish a plurality of support vector machine prediction models, introduce loss function and the correlation coefficient to adjust sample weight, and finally combine the plurality of prediction models as a total prediction model; and utilize the total prediction model to predict a test sample. In the invention, the previous steps are about data pre-processing, which aims to extract useful information in flood datum and eliminate disturbance of redundant information to the forecast; in the third step, the Boosting algorithm is introduced into the flood forecast so as to try to extract a sample of one model that can't learn well for training the next model; in this way, the accuracy of flood forecast can be improved effectively by the combined model; and the last step is used for testing the model effect. According to the experimental datum, the forecast accuracy can be improved effectively by the technical solution.

The invention provides a sample weight allocation method. The method comprises the following steps: acquiring training samples, wherein the training samples include a positive sample set and a negative sample set; calculating the distance of each positive sample couple in the positive sample set, and the distance of each negative sample couple in the negative sample set; determining distance distribution of the positive sample set according to the distance of each positive sample couple in the positive sample set; determining distance distribution of the negative sample set according to the distance of each positive sample couple in the negative sample set; and determining weight distribution of the training samples based on the distance distribution of the positive sample set and the distance distribution of the negative sample set. The invention further provides a model training method, electronic equipment and a storage medium. According to the sample weight allocation method disclosed by the invention, the weight of the sample couples with wrong classification can be increased, and contribution of the samples with classification errors to targeted loss is increased in the modeltraining process, so that model parameters can be well corrected, and the expression ability of the model parameters is improved.

The present invention is generally directed to various methods and systems for adaptive metrology sampling plans that may be employed to monitor various manufacturing processes. In one example, the method comprises creating a plurality of metrology sampling rules, assigning each of the metrology sampling rules a sampling weight value, identifying at least one workpiece that satisfies at least one of the metrology sampling rules, assigning the sampling weight value for each of the satisfied metrology sampling rules with the identified workpieces that satisfy the rules, and indicating a metrology operation should be performed when a cumulative total of the sampling weight values is at least equal to a pre-established trigger value. In further embodiments, the method involves indicating a metrology operation should be performed when a cumulative total of the sampling weight values for one of the metrology sampling rules is at least equal to a pre-established trigger value or indicating a metrology operation should be performed when a cumulative total of the sampling weight values for one of the workpieces is at least equal to a pre-established trigger value.

The invention discloses a method and a system for intrusion detection based on non-negative matrix factorization under sparse representation. The method includes: acquiring network data and host data, and obtaining a level-one audit privilege program of original network data; preprocessing the network data and the host data, and generating network characteristic data and short-sequence vectors; performing non-negative matrix iterative factorization for a data test matrix, and performing sparse representation for a basis matrix and a weight matrix; sampling weight matrix data subjected to sparse representation by the aid of a projection matrix so that highly characteristic weight coefficient vectors are obtained; and matching the highly characteristic weight coefficient vectors with characteristic vectors in training data by the aid of characteristic vector library data, and judging whether abnormal characteristics are conformed to or not. The method and the system for intrusion detection achieve data dimension reduction by non-negative matrix factorization and uses multi-divergence as a measurement level, an RIP (routing information protocol) condition in sparse representation is added into a combined divergence objective function family to restrain a non-negative matrix factorization iterative process, data detection dimensionality is lowered, and high-dimensional mass data processing of the system for intrusion detection is facilitated.

The invention relates to a power system transient state voltage stability evaluating method based on misclassification cost classified learning, which belongs to the power system stability analyzing and evaluating field. Based on dynamic measurement datum of a synchronous phasor measuring unit, a key subsequence closely related to the power system state is extracted from a time sequence constituted of a large amount of dynamic measurement datum; by setting different misclassification costs of stable power system and unstable state, a weight coefficient is introduced into a learning sample; a decision-making tree algorithm infused with the sample weight coefficient is used to perform classified learning so as to acquire a decision-making tree model; and the decision-making tree is used for online monitoring to implement evaluation to the power system transient state voltage stable situation.

The invention provides a multimodality medical image fusion method based on multiscale anisotropic decomposition and low rank analysis. The method comprises the following steps that (1) an image pyramid is established for input images, the images on each layer are subjected to meshing, an anisotropic heat kernel related to data is established, and multiscale representation of the images is achieved; (2) the images in different scales are grouped, low rank analysis is established for each group, low rank parts are extracted, meanwhile, noise is effectively filtered out, and a multiscale space is established by extracted obvious information; (3) in each layer of the image pyramid, low-frequency information is fused by a S-type function, high-frequency information is fused by a maximum selection strategy, and interlayer sampling weights of the pyramid are fused. According to the multimodality medical image fusion method, good robustness is achieved for fusion of noise images.

A micro combined navigate system and a relative self-adaptive filter method belong to inertia navigate system, comprising an abundance distribution inertia measuring unit, a micro GPS receiver, and a navigate computer, wherein the navigate computer comprises a data pick-up, navigate calculation, self-adaptive flat particle filter. And the filter method comprises that first designs a system state equation and a measuring equation, generates state sample particles in prior distribution according to the system state and the self-adaptive theory, then based on UKF filter threshold, judges if the standard decide particles enter into UKF filter, while only refreshes time of the particles not entered into UKF filter, refreshes sample weight, judges if needs resample, processes state fusion on the particles to obtain a state evaluated value. The invention can obtain inertia and navigate property, and process compensation in navigate calculation, while the invention avoids large calculation of flat particle filter, improves the accuracy of navigate system, reduces the error of navigate system, to be used in aviation and aerospace fields or the like.

The invention discloses a weighted least square support vector machine-based soft measurement method for magnetic flux linkage of a bearingless permanent magnet synchronous motor. The method comprises the following steps of: selecting a rotor position angle, torque winding current, levitation force winding current and rotor eccentric displacement of the bearingless permanent magnet synchronous motor as four input variables of a bearingless permanent magnet synchronous motor magnetic flux linkage soft measurement model, wherein the magnetic flux linkage y is taken as an input variable; acquiring representative input variable sample data and output variable sample data, performing normalization preprocessing on both the output variable and the input variables, and forming a modeling sample set for corresponding normalized values; performing residual analysis on the modeling sample set to acquire each sample weight; training the modeling sample set and establishing a bearingless permanent magnet synchronous motor magnetic flux linkage correlation model by using a weighted least square support vector machine; and finally abnormalizing an sy value to acquire the magnetic flux linkage y. Thus, the magnetic flux linkage value of the bearingless permanent magnet synchronous motor is predicted and controlled on line in real time.

The invention discloses a method for reducing dimensions of texture features for surface defect detection on the basis of machine vision. The method has the advantages that noise samples and safety samples among training samples are removed, boundary samples replace randomly selected samples to be used as sample sets used during feature weight iteration, different sample weights are attached to three most adjacent samples according to difference in importance degrees of the three most adjacent samples when the feature weights are computed, accordingly, the pertinence of selection on features with high category correlations is improved, the interference degree of noise is reduced, and the method is high in adaptability; correlation coefficient matrixes are acquired, adaptive threshold values are set, redundant features are removed, the features with the high category correlations are extracted, the dimensions of the features are reduced while the classification and identification accuracy is guaranteed, online feature extraction, classification and identification speeds can be greatly increased, and problems of long online feature extraction time and decrease of prediction accuracy due to the fact that high-dimensional features possibly contain redundant features and even noise features can be solved.

A polyarylene sulfide h as a narrow molecular weight distribution and a high molecular weight and high purity which is industrially useful, wherein the polyarylene sulfide has a weight average molecular weight of 10,000 or more, and weight loss when heated, ΔWr=(W1−W2) / W1×100≦0.18(%) (wherein ΔWr is weight loss ratio (%), W1 is sample weight when arrived at 100° C. and W2 is sample weight when arrived at 330° C.). Its production method includes a polyarylene sulfide prepolymer which contains a cyclic polyarylene sulfide at least 50 wt % or more, and of which weight average molecular weight is less than 10,000 is heated to be converted to a high polymer of which weight average molecular weight is 10,000 or more.

Systems and methods for minimizing extraneous forces and calculating corrected weights of samples based on buoyancy factors for a thermogravimetric analyzer (TGA). The TGA includes a balance chamber and a furnace configured to heat a sample. A null balance is provided in the balance chamber and is used to measure the sample weight change during heating. The furnace includes a cylinder open at the top to receive a sample. The bottom of the cylinder is closed except for a small hole that allows a thermocouple to pass through. An infrared heat source may be provided to heat the cylinder. The balance chamber can be thermally isolated from the furnace using an actively cooled plate and a system of heat shields disposed between the furnace and balance chamber. A thermocouple disk is further provided to limit gas flow in the furnace and increase reliability of sample weight measurements.

The invention discloses an industrial process fault classification method based on self-adaption integrated semi-supervision Fisher discrimination. The method comprises the steps of when off-line modeling is conducted, firstly conducting off-line modeling on unlabeled data, and constituting a semi-supervision random training subset by combining labeled data with the unlabeled data; when iteration training is conducted on a sub classifier each time, conducting semi-supervision Fisher dimensionality reduction to obtain a Fisher discrimination matrix, and obtaining a posterior probability matrix, a combined weight of the sub classifier and a sample weight of the labeled data during next time iteration with the labeled sample data after dimensionality reduction according to a Bayesian statistics method; adopting the posterior probability matrix of the labeled data and a label of the matrix as a training set of a fusion algorithm K near neighbor; during online classification, calling each sub classifier to obtain the posterior probability matrix of an online sample to be detected, and inputting the posterior probability matrix into a fusion K near neighbor classifier with the weight to obtain a final result. Compared with an existing method, the industrial process fault classification method based on the self-adaption integrated semi-supervision Fisher discrimination improves the fault classification result of an industrial process, and more facilitates automated implementation of the industrial process.

The invention provides a training method of classifier, an image detection method and a respective system. The training method is used for training cascading strong classifiers. All strong classifiers can be trained in the following steps of (1) initializing sample weight of all samples according to received numbers of the to-be-trained samples; (2) inputting obtained characteristic values and weight of the samples to a weak classifier for classification trainings so as to minimize the error rate in the weak classifier; (3) based on proportion of bias quantity, updating weight of all samples of the next stage weak classifier according to training results of the current weak classifier; (4) repeating steps (2) and (3) until the last stage of weak classifier is trained; and removing samples in the minimal error category classified by the current stage weak classifier, and inputting other parts into the next stage of weak classifier until the last stage of weak classifier is trained. According to the invention, the trained cascading strong classifiers are used for classifying obtained difference evaluation blocks. Problems of low classification accuracy rate and high training cost are solved.

In a monitoring system for a wood drying kiln having a housing with an interior area for placement of a charge of wood for drying the charge of wood, the interior area having a floor, environment devices for adjusting heat and humidity in the interior area of the kiln, and a control unit for controlling the environment devices, the improvement is a support in the interior area for supporting a representative sample of the charge of wood, a sensor mounted to the support for sensing the weight of the sample which is a function of moisture content in the sample and a proportioning mechanism connected between the sensor and the sample for suspending the sample and for applying a proportioned force resulting from the weight of the sample to the sensor. The sensor is connected to the control unit for generating a signal which is proportional to the weight, and thus, to the moisture content of the sample. A stand on the kiln floor has a member for carrying the support at a location which is spaced from the wood charge without requiring and kiln wall space. The invention also includes the possibility of supporting the sensor from the ceiling and at any location in the kiln whether the sensor is supported from the ceiling or the floor. The invention is also useful in kilns for drying other materials than wood.

The present invention discloses a multi-label AdaBoost integration method based on a label correlation. A stump decision tree method based on a sample weight is used as a weak classifier; an output of the weak classifier is confidence of a sample label; the confidence size closely depends on the sample weight; and the construction method is simple and efficient. According to the present invention, aiming at the mulit-classification problem, the similarity between labels is judged according to a classification result and is fused into the iterative training of a multi-label AdaBoost algorithm; the label correlation analysis is merged into training of a classification model, the label correlation analysis and training of the classification model are mutually promoted and mutually influenced, and finally, the performance of a strong classifier is promoted. Aiming at the multi-label classification problem, a cosine similarity is adopted to calculate a label correlation matrix, an original label is converted to a fuzzy label, and a fuzzy label matrix and classifier model training are combined. The method disclosed by the present invention is easy to implement, the efficiency of a multi-label classification system can be improved, and the method has a better classification effect.

The application discloses a data detection method and a data detection device. The method comprises the following steps: reading a confidence interval determined based on the sample weight of a labeled sample, wherein the sample weight is obtained by training a pre-acquired labeled sample through a weight model; judging whether to-be-checked data is within the confidence interval to get a first judgment result; and judging whether the to-be-checked data is legitimate data according to the first judgment result to get a second judgment result. The technical problem that the accuracy of data legitimacy check is low is solved, and an effect of accurate data legitimacy check is achieved.

The invention provides a measuring method for fluorine ion content in glass. The measuring equipment adapting the measuring method comprises selecting electrode of fluorine ion, reference electrode, PH acidimeter and magnetic stirrer. Reagent comprises sodium hydroxide, phenolphthalein indicator, hydrochloride and regulating buffering solution of total ion intensity. The measuring method comprises following procedures that sample is washed and dried; sample of sodium hydroxide is melted; melted sample is immerged by hot de-ionized water; the regulating and buffering solution of total ion intensity is added; potential value is measured by the selecting electrode of fluorine ion; logarithm value of fluorine ion concentration is found on the calibration curve; fluorine ion concentration is calculated according to the formula of F-%=[(VXCX10) / (GX1000)]X100=(VXC) / G. Where: V--cubage of sample detecting solution (ml); G--sample weight (g); C-logarithm value of fluorine ion concentration found on the calibration curve. The invention is simple, quick and low cost compared with prior technology.

Systems and methods for minimizing extraneous forces and calculating corrected weights of samples based on buoyancy factors for a thermogravimetric analyzer (TGA). The TGA includes a balance chamber and a furnace configured to heat a sample. A null balance is provided in the balance chamber and is used to measure the sample weight change during heating. The furnace includes a cylinder open at the top to receive a sample. The bottom of the cylinder is closed except for a small hole that allows a thermocouple to pass through. An infrared heat source may be provided to heat the cylinder. The balance chamber can be thermally isolated from the furnace using an actively cooled plate and a system of heat shields disposed between the furnace and balance chamber. A thermocouple disk is further provided to limit gas flow in the furnace and increase reliability of sample weight measurements.

The invention provides an oxygen bomb combustion method for testing the content of halogens and the content of sulphur in industrial solid waste. The concentration-spectrum peak area working curve of to-be-measured element series standard solution is drawn, a proper amount of sample is added to an oxygen bomb based on the principle that the amount of sample increases with the reduction of calorific value, absorption liquid the volume of which is 0.05 time that of the oxygen bomb and a proper amount of H2O2 with mass fraction of 10-30% are also added, the sample is burnt and oxidized in the oxygen bomb to obtain a substance which is then absorbed by the absorption liquid so as to generate test sample solution, the spectrum peak area of the sample solution is measured by means of an ion chromatograph, and the content of halogens and the content of sulphur in waste are calculated according to the spectrum peak area and the series standard solution working curve. According to the method, by adding a certain amount of H2O2, oxidation of sulphur in the combustion process can be promoted, complete combustion of sulphur during testing is ensured, loss is reduced, and test results are more accurate; meanwhile, the weight of the sample added each time is determined based on the principle that the amount of sample increases with the reduction of calorific value, so that complete combustion of the sample during testing is guaranteed and testing time is shortened.