55 results about "Eigenvalue distribution" patented technology

The invention discloses a commodity purchase prediction modeling method. The method comprises the steps that a purchase record marking training sample is used to predict whether to purchase or not; a sliding window commodity purchase sample is constructed; commodity purchase features are designed based on a time preference; a gradient improvement decision tree algorithm is used for training prediction; after the sample and the features are constructed, feature processing and selection need to be performed, and then the features are input into the gradient improvement decision tree algorithm for training prediction; and feature selection indicators include feature value distribution and relevancy, feature information gains, feature calling frequency, influences of feature knockout, etc. Ordering is performed on feature importance by integrating the indicators, and redundant features with low importance are eliminated. According to the method, a sliding window sample construction method and a feature system based on the time preference are proposed, the accuracy of a commodity purchase prediction model is effectively improved, and the method is used for realizing commodity personalized recommendation in a big data background to precisely recommend proper commodities to a user at a proper time and a proper place.

The present invention discloses an MMC sub module open-circuit fault diagnosis method. The method comprises the steps of: constructing a random matrix based on a sub module capacitive voltage value, employing a monocyclic theorem to analyze the distribution of feature values of the random matrix to determine the operation state of the MMC; if the sub module open-circuit faults are detected, performing statistic analysis of the sub module capacitive voltage of a fault bridge to locate a fault sub module. The MMC sub module open-circuit fault diagnosis method can be compatible with single / multi-sub module open-circuit fault diagnosis, the used sub module capacitive voltage value has been existed in the current MMC control system so as not to need extra hardware resources. Besides, the faultdiagnosis is completed based on data analysis without depending on an MMC analysis model so as not to be sensitive to the nondeterminacy of system parameters, and therefore, the MMC sub module open-circuit fault diagnosis method has high robustness.

A subject information obtaining device according to the present disclosure includes: a first optical characteristic value distribution obtaining unit configured to obtain a first optical characteristic value distribution based on a first detection signal obtained by detecting first photoacoustic waves generated by irradiating light in a first measurement state on a subject; a second optical characteristic value distribution obtaining unit configured to obtain a second optical characteristic value distribution based on a second detection signal obtained by detecting second photoacoustic waves generated by irradiating light in a second measurement state different from the first measurement state on the subject; and a data processing unit configured to obtain a similarity distribution based on the first optical characteristic value distribution and the second optical characteristic value distribution.

There is disclosed an echo cancelling apparatus in a digital mobile communication system and method thereof, which is incorporated in a single chip together with a speech coder, and can improve convergence rate by using an excited signal as an input to n adaptive filter and also obtain a stabilized convergence characteristic while using a NLMS algorithm. The echo canceller according to the present invention uses the excited signal obtained at the speech coder as an input signal to an adaptive filter thereby to lower the Eigenvalues distribution rate of the speech signal, thus providing an improved performance compared to the conventional echo canceller.

The invention discloses a tractor-network coupled system stability criterion calculation method based on an impedance return ratio matrix. A high-speed rail vehicle network cascade system is used as arepresentative MIMO cascade system. The method includes the following steps: deducing the equivalent front-stage output impedance of the system; deducing the equivalent back-stage input impedance ofthe system; calculating the overall transfer function of the system, and getting a return ratio matrix and a transposed matrix thereof; estimating the feature value of the return ratio matrix and thefeature value of the transposed matrix of the return ratio matrix, and limiting the distribution area; and setting a forbidden zone for the distribution area of the feature values, and getting a lessconservative MIMO system stability criterion. The criterion can be used to simply and effectively analyze the stability of the MIMO cascade system, and is much less conservative than the existing singular value criterion, norm criterion or the like.

The invention discloses an error evaluation method and system for a non-stationary output voltage transformer. The method comprises the following steps: representing an evaluation parameter of an error state of the voltage transformer through employing a differential process under a non-stationary output condition; evaluating the evaluation parameters of the error state of the voltage transformerin real time in combination with a high-dimensional random matrix and an M-P law, and evaluating the error state of the voltage transformer, wherein the error state comprises an error normal state andan error abnormal state; and if the voltage transformer is in the error abnormal state, judging the error polarity of the voltage transformer in the error abnormal state based on the correlation between in-phase voltage signals. Real-time evaluation of the error state and the error polarity can be realized only according to the non-stable output data of the voltage transformer, so that the evaluation cost can be greatly reduced, and the operation and maintenance level of the voltage transformer can be enhanced. Compared with a circular ring law, the M-P law adopted in the high-dimensional random matrix has the following advantages: the eigenvalue distribution is easier to solve, and the evaluation index sensitivity is higher.

The invention belongs to the technical field of situation awareness for a dynamic electric power system, and discloses a recursive state estimation fused anomaly detection method for the dynamic electric power system. The method comprises the following steps: firstly establishing a simplified non-linear recursive model of system node voltage according to the dynamic characteristic of the electricpower system, and reasonably representing the influence of the dynamic change of electric power system load on system node voltage; then realizing system node voltage dynamic estimation based on the non-linear recursive model based on a recursive state estimation filtering algorithm, and on this basis, further constructing a residual error random matrix of system node voltage; finally constructinga system residual error dynamic performance index based on characteristic spectrum mean value-square deviation statistical analysis so as to effectively reflect the influence of electric power systemanomaly on the residual error matrix eigenvalue distribution of system node voltage, and then judging according to a self-adaptive statistical threshold value to finally realize effective state estimation and anomaly detection of the dynamic electric power system.

The invention provides an efficient pretreatment method for special saddle point problems. The method includes the steps that according to the structural feature of the saddle point problem (3), structured augmentation-free and Schur complement-free block triangle preconditioners (11) and (12) are designed; spectral analysis is conducted on the designed preconditioners, and it is analyzed that a preconditioning matrix can have better feature value aggregation when appropriate parameters are selected; meanwhile, the defect that an original block opposite angle preconditioner can only solve the problem that k<2> is smaller than 1 is overcome, and the application range is widened; according to the structure feature of the asymmetrical saddle point problem (6) with a singular (1,1) block, a more generalized structured block triangle preconditioner is designed; spectral analysis is conducted on the designed preconditioner (13), the feature value distribution of the designed preconditioning matrix, the corresponding feature vector and the minimal polynomial are studied, and the feature value aggregation when the optimal parameter is selected is analyzed.

The invention discloses an MIMO cascaded system stability analysis method based on an impedance return-ratio matrix. The MIMO cascaded system stability analysis method regards a high-speed rail train network cascaded system as an object, and comprises the steps of deducing and calculating equivalent pre-stage output impedance of the system, deducing equivalent post-stage input admittance of the system, calculating a system integral transfer function and acquiring a return-ratio matrix, estimating feature values of the return-ratio matrix and limiting a distribution region of the feature values, setting a forbidden region in the feature value distribution region, and acquiring MIMO system stability criterion with low conservative property. The deduction and calculation of the equivalent pre-stage output impedance and the equivalent post-stage input admittance of the system are the premises for acquiring the system integral transfer function, and are essential conditions for further acquiring the return-ratio matrix of the cascaded system; and the Gerschgorin circle theorem is the basis for setting the forbidden region to limit and avoid the matrix feature value range, and further reducing the conservative property of the criterion. The criterion method is suitable for stability analysis of the MIMO cascaded system, can analyze stability of the MIMO cascaded system simply and effectively, and has lower conservative property than that of the existing singular value criterion, norm criterion and the like.

The invention provides a data re-deletion method and storage equipment. The method is applied to storage equipment in a distributed storage system. The method comprises the following steps of: responding to-be-written data in an IO request and calculating a feature value of the to-be-written data; when the feature value of the to-be-written data is sent to another storage equipment according to a feature value distribution strategy, receiving a response message returned by the another storage equipment through carrying out search on the basis of the feature value of the to-be-written data; and carrying out re-deletion processing on the to-be-written data according to the response message. The method is capable of reducing the network load of the distributed storage system and reducing the probability that IO reading needs to carry out data transmission through a network; and the transmission load of a data network is reduced, so that the delay of IO reading and writing is shortened and the IO processing speed and throughput capacity are enhanced.

The invention discloses a near-field accompanying optimization method for far-field sonic boom suppression of a supersonic aircraft, and the method comprises the steps: giving a parameterized initialnear-field waveform, obtaining near-field and far-field overvoltage signals, and carrying out the modal decomposition; extracting each decomposition mode and characteristic value distribution thereof,and reversely solving a near-field overvoltage signal; on the basis of CFD grid division, according to the reversely solved near-field overvoltage signals and a given line segment at the lower position of the fuselage, completing target function variation formed by near-field overvoltage and target overvoltage distribution, and constructing and solving an adjoint equation; and performing optimization design based on a sequential quadratic programming algorithm, and performing forward calculation check on a design result. According to the method, the defects of traditional near-field accompanying optimization can be overcome, meanwhile, the tedious operation of far-field-near-field-body variation of far-field signal design is avoided, and the efficiency of accompanying optimization can becontinuously and fully utilized.

The invention provides a textile defect detection method based on a peak value threshold, rotation calibration and a mixed characteristic. The method is characterized by analyzing digital image pixelgray scale information of a flat textile surface under an illumination light source; segmenting an image into grids which are not overlapped; calculating IRM, HOG, GLCM and Gabor characteristic valuesof each grid; and according to characteristic value distribution, automatically positioning a textile surface defect. The method is especially suitable for automatically identifying the textile surface defect in a textile flat surface gray scale digital image collected under a stable illumination light source.

The invention discloses a multivariate flame monitor-based on-line judgment method for fuel type, which comprises the following steps: when a known fuel and a new fuel exist in burning, extracting characteristic values in a time domain and a frequency domain from flame radiation signals of the known fuel, and mathematically transforming the characteristic values to acquire orthogonal characteristic values; establishing a joint probability density model and a neural network model of characteristic value distribution of the known fuel based on the orthogonal characteristic values; extracting characteristic values in the time domain and the frequency domain from burning flame radiation signals of a fuel to be identified, mathematically transforming the characteristic values to acquire orthogonal characteristic values and inputting the orthogonal characteristic values into the joint probability density model of various known fuels to judge, and if the fuel to be identified in burning is the new fuel, storing the orthogonal characteristic values of the flame radiation of the new fuel to establish the joint probability density model of the new fuel and update the neural network model; and if the fuel to be identified in burning is not the new fuel, judging the type of the fuel through the neural network model.

The present invention discloses a characteristic selection method based on sample characteristic distribution confusion degree, which comprises the following steps: 1) sorting the value sets of each characteristic fi of all kinds of samples in the data set X in an ascending order; 2) determining the range of the limited value of the characteristic fi corresponding to each kind of the sample; 3) determining the number of confusing samples corresponding to the ith characteristic of the M-kind sample in the data set, and calculating the characteristic distribution confusion degree of the ith characteristic in the data set X so as to use the same method to obtain the Confusion value of each characteristic in the data set X; 4) ranking the characteristics in the data set X according to their significance and the obtained Confusion values to obtain an ordered characteristic set F; and 5) based on the set subset search strategy, using a classifier to search ordered characteristic set F or the subset Fsub formed by the partial characteristics in the ordered feature set F to obtain a desired characteristic subset D. The method can select the better-performance characteristic subsets, which improves the recognition capability for the characteristic subsets, and reduces the search times in the subset searching process.

The invention discloses a fresh fuel online identification method comprising two stages: in the first stage, establishing a joint probability density model of known fuel characteristic value distribution; extracting characteristic values of a flame radiation signal in a time domain and a frequency domain from a known fuel burning flame radiation signal, and taking the characteristic values as flame original characteristic values; changing the flame original characteristic values into mutually uncorrelated orthogonalized characteristic value data through orthogonalized data processing; utilizing the obtained orthogonalized characteristic value data to establish the joint probability density model of each known fuel characteristic value distribution; in the stage two of fresh fuel online identification, extracting characteristic values of a to-be-identified fuel burning flame radiation signal in the time domain and the frequency domain, and taking the characteristic values as original characteristic values; carrying out orthogonalized data processing on the original characteristic values; inputting the obtained orthogonalized characteristic value data to the joint probability densitymodel of each known fuel characteristic value distribution so as to obtain a probability density value that the to-be-identified fuel belongs to each known fuel, and judging the type of the to-be-identified fuel according to the probability density value and judging whether the to-be-identified fuel is a new fuel or not.

The invention relates to a grid stability judging method based on grid PMU big data, and belongs to two fields of grid big data analysis and grid stability analysis. The method comprises the following steps: firstly, performing modeling according to grid PMU data through a random matrix; secondly, analyzing grid PMU historical data, and establishing a distribution up and down boundary map of the PMU historical data; thirdly, analyzing grid PMU real-time data, and establishing a characteristic value distribution map of the PMU real-time data; fourthly, displaying and comparing the characteristic value distribution map of the PMU real-time data and the distribution up and down boundary map of the PMU historical data on the same interface, so as to judge whether a grid is stable or not. According to the method, the grid stability is cognized according to statistic characteristics, such as the PMU historical data and the PMU real-time data, and the requirements of the grid, and can be displayed through a human-computer interface, so that the grid is convenient for a grid operation and maintenance administrator to use.

A resistor array is provided in an element array. A mean value of a characteristic-value distribution is associated with a median of combined resistance values obtained by the element array. An array of trimming information pieces corresponding to combined resistance values larger than the median is set in a descending order of ‘15’ to ‘8’ in decimal number, and an array of trimming information pieces corresponding to combined resistance values less than the median is set in an ascending order of ‘0’ to ‘7’ in decimal number. A circuit converts trimming information derived from the trimming information generation circuit to generate element selection information for selecting turn-off resistors to obtain combined resistance values from the resistor array. Thus, the number of melted-and-cut fuses involved in generation of trimming information associated within the range of “mean value±2σ” in the distribution is reduced.

The invention discloses a nonlinear system characteristic variable significance mining method based on data stream processing, wherein the method comprises the following steps: collecting the historical data and preprocessing the historical data; DBSCAN parameters are optimized by closed-loop structure, and the data set is divided by DBSCAN clustering. A nonlinear model based on LGBMRegressor is established for each type of data, and the number of times each feature variable is used in the iterative training process of the model is recorded to characterize its significance, and the sequence ofsignificant feature variables is established. Determine the main distribution intervals of each variable value, form the eigenvalue distribution interval model of a single class, integrate the eigenvalue distribution interval models of all classes to establish the grid model, and map the salient eigenvariable series to the corresponding class in the grid model to form the composite grid model; Build the real-time computing framework based on Storm, design the Topology (topology) based on stream processing, and load the composite mesh model at the same time.

The invention discloses a power grid mountain fire potential fire point identification method based on a two-dimensional OTSU. The method comprises the following steps: firstly, dividing a satellite image data plane into a plurality of rectangular sub-regions according to latitude and longitude; Selecting a sub-region as a to-be-identified region, calculating an average channel brightness temperature characteristic value of pixel points in a neighborhood of each pixel point in the to-be-identified region, and calculating two-dimensional probability distribution of the pixel brightness temperature characteristic values in the to-be-identified region; And obtaining a fire point determination optimization threshold based on a two-dimensional OTSU method, and determining whether each pixel point in the to-be-identified area is a potential fire point according to the optimization threshold. According to the method, the judgment threshold of the potential fire point is calculated only according to the current image; According to the method, the characteristic value distribution information of the whole pixels in the region is fully utilized, the adaptive adjustment of the mountain fire judgment threshold can be carried out according to different time and different regions without the help of historical threshold values, the alarm accuracy is high, the calculation efficiency is high,the practicability is high, and the method plays an important role in maintaining the safety and stability of a power grid and preventing and reducing disasters.

The invention relates to a data processing method and device based on artificial intelligence, computer equipment and a storage medium. The method comprises the steps of obtaining a preset number of samples, wherein the samples comprise user features and user behavior sequences; performing sequence pattern mining based on the user behavior sequence to obtain a plurality of user behavior sequence patterns; identifying the relevancy between the user characteristics and the user behavior sequence mode, and extracting a strong correlation sequence mode corresponding to the user characteristics according to the relevancy; counting characteristic value distribution corresponding to each user characteristic in the strong correlation sequence mode; and when one sample has the user feature with themissing feature value, filling the missing feature value according to the feature value distribution to obtain a sample filled with the feature value. A sample containing accurate and effective userfeatures can be provided for model learning, and then the accuracy of machine learning is improved.

A base station apparatus, in a wireless communication system supporting band aggregation, able to limit drops in throughput while lowering the amount of control information for the rank of an added band; and a terminal device and rank setting method of the same. An added band rank setting unit (108) includes a memory (1084) which defines, as a rank and from the eigenvalue distributions of the channel matrices of the master band and added band, the number of eigenvalues of the channel matrix of the added band that satisfy channel quality according to the rank of the master band, and associates that number with the frequency band and rank of the master band and the frequency band of the added band. The added band rank setting unit (108) acquires, from the memory (1084), the rank that is associated with the information on the frequency band and rank of the master band and the information on the frequency band of the added band, and sets the same as the rank of the added band.

An engine fuel-injection quantity abnormal fault diagnosis method based on information fusion mainly requires a data collecting system, a fuzzy processing system and a neural network system. The data collecting system collects and fuses air flow sensor signal waveform characteristics of an engine and storage battery voltage waveform characteristics, and the characteristics are introduced into the fuzzy processing system. The fuzzy processing system fuzzifies the characteristics and enables characteristic values to be distributed in a range of [0,1], and the data is applied to the neural network system. The neural network system is a back propagation (BP) neural network of a multi-output structure, an input layer is a fuzzified waveform characteristic value, and an output layer is diagnosed faults. The engine fuel-injection quantity abnormal fault diagnosis method has the advantages of fusing information of automobile multiple components, improving abnormal fault diagnosis accuracy of the engine fuel injection quantity through correlation among components, and providing technical assistance for automobile maintaining enterprises. The engine fuel-injection quantity abnormal fault diagnosis method can be widely applied to diagnosis of various automobile faults.

The invention provides a textile defect detection method based on the peak threshold and mixed characteristics. According to the method, the digital image pixel gray information of the flat textile surface based on an illumination light source is analyzed, an image is divided into non-overlapped grids, IRM, HOG, GLCM and Gabor characteristic values of each grid are calculated, and textile surfacedefect is automatically positioned according to characteristic value distribution. The method is advantaged in that the method is especially suitable for automatically identifying the textile surfacedefect of the gray digital image of the flat textile surface acquired under the stable light source illumination condition.

The invention provides an advertisement recommendation model monitoring method and device, an advertisement recommendation method and device, electronic equipment and a computer readable storage medium. The method comprises the steps of for two adjacent model updating periods, obtaining a user sample which can be used when an advertisement recommendation model is trained in each model updating period to recommend an advertisement; determining a characteristic value quantile of a user characteristic type corresponding to the user sample; performing distribution estimation processing according to the eigenvalue quantiles to obtain eigenvalue probability distribution of the user feature types corresponding to each model updating period; measuring the difference between the characteristic value probability distributions of two adjacent model updating periods to obtain information divergence corresponding to the user characteristic type; and according to the information divergence, selecting a user sample from the user samples corresponding to the two adjacent model updating periods as a training sample. Through the method and the device, the sensitivity to characteristic value distribution fluctuation can be improved, the exception can be captured more accurately, and the advertisement recommendation precision is improved.

The invention discloses a dimension reduction and correlation analysis method suitable for large-scale data, and the method comprises the steps: projecting high-dimensional data to a Fourier domain, and converting a solving feature vector problem of linear correlation analysis into a meaningful Fourier domain basis; because the Fourier domain basis is predefined and the characteristic value distribution of the data is ordered, accelerating training by inputting the training samples in batches until the required Fourier basis is stable and ordered; determining the number of Fourier bases and aprojection matrix, and multiplying the projection matrix by the high-dimensional data set to obtain a low-dimensional data set so as to facilitate rapid processing of data. According to the data dimension reduction method, on the basis of fast Fourier transform and correlation analysis, noise and redundant information in a high-dimension data set can be removed, unnecessary operation processes indata processing are reduced, and the running speed and the memory use efficiency in data dimension reduction calculation are improved.

The invention provides a textile defect detection method based on a peak coverage value and mixed features. According to the method, pixel grayscale information of a digital image of a flat textile surface under an illumination light source is analyzed, the image is segmented into grids which are not mutually overlapped, IRM, HOG, GLCM and Gabor feature values of each grid are calculated, and a textile surface defect is automatically positioned according to the distribution of the feature values. The textile defect detection method is especially applicable to automatically recognizing textilesurface defects in textile flat surface grayscale digital images collected under a stable illumination light source.

A resistor array is provided in an element array. A mean value of a characteristic-value distribution is associated with a median of combined resistance values obtained by the element array. An array of trimming information pieces corresponding to combined resistance values larger than the median is set in a descending order of ‘15’ to ‘8’ in decimal number, and an array of trimming information pieces corresponding to combined resistance values less than the median is set in an ascending order of ‘0’ to ‘7’ in decimal number. A circuit converts trimming information derived from the trimming information generation circuit to generate element selection information for selecting turn-off resistors to obtain combined resistance values from the resistor array. Thus, the number of melted-and-cut fuses involved in generation of trimming information associated within the range of “mean value±2σ” in the distribution is reduced.

The invention discloses an evaluation method and device for face image similarity, and the method comprises the steps: collecting face image information, obtaining a face multi-dimensional feature vector, and carrying out the In < t > quantification of the face multi-dimensional feature vector; carrying out dimensionality reduction on the quantized face multi-dimensional feature vectors, and taking d-dimensional feature values with high contribution degrees; training face image data, counting feature value distribution of each dimension and segmenting to divide a value domain of the feature value of each dimension into k sections with unequal widths; for the two face image feature vectors of which the similarity is to be recognized, obtaining the similarity of each dimension according to the segment where the feature value of each dimension is located, and obtaining an evaluation value of the similarity of the face image vectors according to the similarity of each dimension. Accordingto the embodiment of the invention, the method can reduce the consumption of system resources, achieves the quick and objective calculation of the similarity of the feature vectors of different humanface images, improves the efficiency and applicability of a clustering algorithm, and can achieve the quick and accurate judgment of the similarity of the feature vectors of the human face.

The embodiment of the invention provides an unmanned aerial vehicle rigid formation switching method and device, and the method comprises the steps: obtaining at least one corresponding feasible topological graph based on a current formation topological graph; screening the at least one feasible topological graph based on a distribution rule corresponding to the feature value of the feasible topological graph incidence matrix to obtain at least one first target topological graph; based on the obstacle model, calculating the minimum effective radius and the expansion factor of each first targettopological graph; and based on an energy optimization algorithm, obtaining a topological graph with the minimum cost from the first target topological graphs as a second target topological graph, and achieving rigid formation switching by the unmanned aerial vehicle based on the second target topological graph. According to the characteristic value distribution rule of the feasible topological graph, the solution space of the feasible topological graph is greatly reduced; by establishing the effective radius of the formation, the formation is ensured to smoothly pass through an obstacle, andthe physical feasibility of formation change is improved; and through an energy optimization algorithm, the minimum formation change cost is ensured.