82 results about "Alternating least squares" patented technology

A method of multivariate spectral analysis, termed augmented classical least squares (ACLS), provides an improved CLS calibration model when unmodeled sources of spectral variation are contained in a calibration sample set. The ACLS methods use information derived from component or spectral residuals during the CLS calibration to provide an improved calibration-augmented CLS model. The ACLS methods are based on CLS so that they retain the qualitative benefits of CLS, yet they have the flexibility of PLS and other hybrid techniques in that they can define a prediction model even with unmodeled sources of spectral variation that are not explicitly included in the calibration model. The unmodeled sources of spectral variation may be unknown constituents, constituents with unknown concentrations, nonlinear responses, non-uniform and correlated errors, or other sources of spectral variation that are present in the calibration sample spectra. Also, since the various ACLS methods are based on CLS, they can incorporate the new prediction-augmented CLS (PACLS) method of updating the prediction model for new sources of spectral variation contained in the prediction sample set without having to return to the calibration process. The ACLS methods can also be applied to alternating least squares models. The ACLS methods can be applied to all types of multivariate data.

The invention discloses a personalized recommending method fused with user trust relationships and comment information. The method includes adding useful comments and user trust relationships on the basis of performing probability decomposition on a score matrix, wherein the user trust relationships explicitly show the trust level of users, the user useful comment information potentially shows the trust level of the users, and the two aspects of information can predict interests and hobbies of the users; and an alternating least squares is used to train model parameters. The personalized recommending method can fuse user trust relationships in a credible network and the potential trust relationships acquired by the useful comment behaviors, and can improve the recommendation precision.

The invention provides a method for classifying the rail failures of a high-speed rail. The main idea is that the method comprises the steps of extracting local features of a time domain and a frequency domain of damaged signals by using a wavelet analysis method; building a three-dimensional tensor signal for a same measuring point by combining different compartments; expanding data to a multi-dimensional space to obtain a non-negative tensor; taking an alternate least squares algorithm as an iteration criterion of the non-negative tensor decomposition; introducing SVD (Singular Value Decomposition) to improve the initialization of the non-negative tensor; extracting hidden features by an improved non-negative tensor decomposing method; and finally, introducing an extreme learning machine algorithm to realize real-time classification on the rail failures. According to the method for classifying the rail failures of the high-speed rail provided by the invention, the signals of rail defects and failures can be classified accurately, the classifying speed and accuracy of the for classifying the rail failures can be improved, and the robustness can be realized; furthermore, the classifying method based on the g the rail failures is prior to an existing method, the better recognition effect can be obtained, and the method can be extensively applied to the field of classifying the for classifying the rail failures.

The invention relates to a method for personally recommending software projects for open source communities. TF-IDF (term frequency-inverse document frequency) characteristics of project contents are extracted and are combined with known project evaluation of users, the similar projects which are combined with personality and project requirements of the users are recommended, candidate recommendation results are computed for the personality of the users by means of ALS (alternating least squares) collaborative filtering, each project characteristic is computed by the aid of term frequencies-inverse document frequencies, candidate recommendation results are computed according to the similarity of the project characteristics, recommendation results for the personality of the users and recommendation results for the project characteristics are linearly combined with one another and are filtered and ranked, and accordingly recommendation results can be generated. The method has the advantages that shortcomings of incapability of completely embodying project characteristics, coarse granularity, low accuracy and the like of existing recommendation technologies can be overcome by the aid of the method; the similar projects for the personality of the users and the project contents can be recommended from angles of known project scores of the users and known project contents, accordingly, the project contents similar to projects currently developed by the users can be recommended, the method has personal recommendation functions and can bring reference or reuse convenience for the users, and the project searching efficiency of developers can be improved.

Several full-spectrum imaging techniques have been introduced in recent years that promise to provide rapid and comprehensive chemical characterization of complex samples. One of the remaining obstacles to adopting these techniques for routine use is the difficulty of reducing the vast quantities of raw spectral data to meaningful chemical information. Multivariate factor analysis techniques, such as Principal Component Analysis and Alternating Least Squares-based Multivariate Curve Resolution, have proven effective for extracting the essential chemical information from high dimensional spectral image data sets into a limited number of components that describe the spectral characteristics and spatial distributions of the chemical species comprising the sample. There are many cases, however, in which those constraints are not effective and where alternative approaches may provide new analytical insights.For many cases of practical importance, imaged samples are “simple” in the sense that they consist of relatively discrete chemical phases. That is, at any given location, only one or a few of the chemical species comprising the entire sample have non-zero concentrations. The methods of spectral image analysis of the present invention exploit this simplicity in the spatial domain to make the resulting factor models more realistic. Therefore, more physically accurate and interpretable spectral and abundance components can be extracted from spectral images that have spatially simple structure.

A spectroscopic analysis method in which spectral data of mixtures obtained from a plurality of points on a sample surface are resolved into component spectra and concentrations. A new alternating least squares multivariate curve resolution technique is presented which iteratively resolves the components. The technique starts from an initial estimate that the spectral values of a first component of the sample are all equal (an ‘empty model’), and resolves that component. Then successive further components are iteratively resolved, from initial ‘empty model’ estimates of those components and from previously resolved spectra. In the common case where the main component is present in nearly pure form in the data set, this empty modelling technique results in more accurate resolution of the components. This is due to the ability of the technique to resolve the pure spectra of minor components without modelling concentrations of the main component into them.

The invention discloses a collaborative filtering recommendation method based on search behavior perception. The method includes the following steps: (1) analyzing inquiry behaviors and search behaviors of a user on an e-commerce website; (2) constructing user-product keyword tensor based on search behavior context; (3) constructing a training dataset of a factor machine model and mapping tensor data into vector data; (4) establishing a recommendation method frame based on the factor machine model and utilizing an improved alternate least squares algorithm to carry out parameter estimation; and (5) evaluating the recommendation method based on the factor machine model through an experiment. The collaborative filtering recommendation method based on the search behavior perception utilizes context information ignored by a traditional collaborative filtering recommendation method, solves the problem that a traditional individuation recommendation method cannot provide a recommended result which expresses the intent of the user on a business to business e-commerce website, and is better than a traditional method in accuracy and timeliness of the recommended result.

Several full-spectrum imaging techniques have been introduced in recent years that promise to provide rapid and comprehensive chemical characterization of complex samples. One of the remaining obstacles to adopting these techniques for routine use is the difficulty of reducing the vast quantities of raw spectral data to meaningful chemical information. Multivariate factor analysis techniques, such as Principal Component Analysis and Alternating Least Squares-based Multivariate Curve Resolution, have proven effective for extracting the essential chemical information from high dimensional spectral image data sets into a limited number of components that describe the spectral characteristics and spatial distributions of the chemical species comprising the sample. There are many cases, however, in which those constraints are not effective and where alternative approaches may provide new analytical insights.For many cases of practical importance, imaged samples are “simple” in the sense that they consist of relatively discrete chemical phases. That is, at any given location, only one or a few of the chemical species comprising the entire sample have non-zero concentrations. The methods of spectral image analysis of the present invention exploit this simplicity in the spatial domain to make the resulting factor models more realistic. Therefore, more physically accurate and interpretable spectral and abundance components can be extracted from spectral images that have spatially simple structure.

This disclosure provides methods, apparatus and systems for performing image processing regression for approximating multidimensional color transformation. According to an exemplary method, a shaping matrix is selected to minimize a cost function associated with a local linear regression representation of the color transformation. In addition, an alternating least squares algorithm is utilized to jointly optimize regression and shaping parameters.

The invention provides an online high-speed railway steel rail damage monitoring method. The method comprises the steps of installing acceleration sensors along a high-speed railway track according to a given distance, acquiring a vibration signal of the steel rail, and forming a sensor network; judging whether the damage exists or not by utilizing a processor on each sensor node, transmitting a damage signal to an information center or a flaw detector to issue an alarm or to be further processed by virtue of the sensor network if the damage exists, and the method is characterized in that the method for judging the damage is based on sparse non-negative matrix factorization characteristic extraction and support vector machine classification, the sparse non-negative matrix factorization adopts singular value decomposition to initialize a matrix, and the iterative computation is carried out by utilizing an alternating least squares algorithm. By adopting the method, an accurate high-speed railway steel rail monitoring result can be acquired, and the damage judgment speed and the damage judgment accuracy can be improved. The method can be widely used for monitoring the damage of the steel rail.

The invention discloses a fast measuring method of tobacco shred mixing ratio, which takes a light absorbency value corresponding to a 5500 to 4200cm near-infrared spectrum of a diffuse reflection near-infrared spectrogram of finished tobacco shred samples to be measured, a diffuse reflection near-infrared spectrogram of leaf-silk samples to be measured, a diffuse reflection near-infrared spectrogram of cut stem samples to be measured and a diffuse reflection near-infrared spectrogram of slice silk samples to be measured as an independent variable matrix; the maxing ratio of leaf silks, cut stems and slice silks used for producing the finished tobacco shreds forms a dependent variable matrix; and a multivariate curve resolution-alternating least square method is used for calculating the matrix in the third step, thus obtaining the actual mixing ratio of the leaf silks, the cut stems and the slice silks in the tobacco shred samples to be measured. The method overcomes the defects of relatively complicated model building process, relatively small model application range, large work load and the like existing in the prior art, is a method that uses less measuring instruments, has simple process and can realize fast onsite measurement, and can be widely applied to the tobacco industry.

The present invention discloses a power generation process control system fault detection method, which comprises the steps of: carrying out matrix factorization on a training matrix X after noise reduction and standardization processing by adopting a principal component analysis PCA method, and adopting a score matrix T as an initial value W0 of a base matrix W; carrying out iteration solution on the base matrix W and a weight coefficient matrix H of the training matrix X by adopting an alternating least square method with nonnegative constrains; constructing a monitoring statistics Tn <2> and SPEn based on nonnegative matrix factorization, calculating probability density functions PDF of the monitoring statistics Tn <2> and SPEn separately by utilizing a kernel density estimation method, setting a significance level and solving control limits of the monitoring statistics Tn <2> and SPEn separately; and calculating to obtain an approximate value W-hat test of a base matrix of a test matrix Xtest by utilizing the weight coefficient matrix H and the test matrix Xtest after data processing, calculating monitoring statistics Tn <2> and SPEn of the test matrix Xtest separately, and indicating that a fault occurs if the monitoring statistics exceed the control limits when compared with the corresponding control limits. The power generation process control system fault detection method can be used for carrying out condition monitoring on massive operation data in the power generation process, further achieve the fault diagnosis of a power generation process control system.

The invention discloses a tensor decomposition model-based interest point type prediction method. The method includes the following steps that: a) the access point coordinate information and stay timeof users in T time periods in each day are obtained from new energy vehicle trajectory data; b) the ranges of the stay areas of the users are defined as uncertain areas according to time periods whenthe users are located at access points, the durations of the stay of the users, the IDs of the types of interest points in the uncertain areas are obtained, and tensors are constructed; c) clusteringprocessing is performed on trajectory data corresponding to all the users; d) the types of the interest points of the users are predicted by using a tensor decomposition model; e) after the information of work, places of residence and places of departure is combined, the similarity of the types of interest points between the places of departure and destinations are adopted as the regularization item constraint of tensor decomposition, and the low-rank approximate solutions of the tensors are calculated; and f) an objective function is iteratively optimized by using an alternating least squares method, and the probabilities of the access of different kinds of interest points by the users are obtained and are adopted as a prediction results. With the method of the invention adopted, the types of the interest points accessed by the users can be accurately predicted.

The invention discloses a method for automatically analyzing a GC-MS (Gas Chromatography-Mass Spectrometry) overlapping peak to accurately recognize compounds. The method comprises the following steps: acquiring data information of compounds in a characterized sample by utilizing GC-MS, and extracting chromatographic peak information in the data information by adopting a multi-scale Gaussian smoothing function aiming at a chromatographic signal (EIC) under each m / z; and with the shape and retention time of the EIC chromatographic peak as the similarity criterion by utilizing a clustering method based on a density function, clustering the EIC peaks belonging to the same compound together, and realizing analysis of overlapping compounds in a single sample by adopting non-negativity constraint multivariate curve resolution-alternating least-squares. The method disclosed by the invention is capable of rapidly and accurately realizing automatic analysis of the compounds in the samples and screening differential compounds and has important value on analysis of complex plant samples, such as non-targeted metabonomic study.

The invention relates to the artificial intelligence, and provides a directional advertisement putting method and device and a computer readable storage medium. The method comprises the steps of calculating the similarity of the label vectors of two users, and obtaining a score 1; calculating the similarity of the two user behavior vectors to obtain a score 2; modeling a user vector by using an alternating least square algorithm, obtaining the user vector and an advertisement vector by using a random gradient descent optimization algorithm, calculating the similarity of the user vectors of twousers to obtain a score 3, and weighting the score 1 to the score 3 linearly to obtain a score 4; carrying out XGBOOST, SVM and GBDT on the user behavior vector, the user interest label vector and the user vector respectively to obtain the corresponding scores as FM input, obtaining a score 5, and weighting the score 4 and the score 5, outputting the final user similarity, comparing the final user similarity with a similarity threshold, and selecting the user higher than the similarity threshold as a target customer group.

The invention discloses a video-on-demand program recommendation method based on a user set. The method comprises the following steps of: 1) collecting the watching records of mass users, setting basic user types, matching the basic user types with film classification labels, and classifying a label weight table for different basic user types; 2) collecting the watching records of the users, carrying out calculation to obtain the model FUser (Li) of a user to be recommended, determining the type of the user to be recommended, and taking a corresponding film classification label weight table as a filtering rule; and 3) collecting the watching records of the users, using an ALS (Alternating Least Squares) algorithm to generate a film recommendation list, and generating a final film recommendation list through the filtering rule. By use of the method, the accuracy and the coverage rate of video recommendation oriented to television end users are high and are unlikely to be affected by popular videos, and a phenomenon that video recommendation information received by the user is repeated is avoided. By use of the method, on the basis of the ALS recommendation algorithm, the characteristics of the television end users are combined to optimize a recommendation result and improve the watching experience of the television end users.

A spectroscopic analysis method in which spectral data of mixtures obtained from a plurality of points on a sample surface are resolved into component spectra and concentrations. A new alternating least squares multivariate curve resolution technique is presented which iteratively resolves the components. The technique starts from an initial estimate that the spectral values of a first component of the sample are all equal (an 'empty model'), and resolves that component. Then successive further components are iteratively resolved, from initial 'empty model' estimates of those components and from previously resolved spectra. In the common case where the main component is present in nearly pure form in the data set, this empty modelling technique results in more accurate resolution of the components. This is due to the ability of the technique to resolve the pure spectra of minor components without modelling concentrations of the main component into them.

The invention discloses a complex field blind source separation method. A complex filed target matrix system is built, and real symmetrization is carried out to obtain a reconstructed target matrix system formed by a real-value target matrix, the complex field combined diagonalization problem is converted into the real field combined diagonalization problem to solve the complex field blind source separation problem; compared with other algorithms that are also suitable for the complex filed, the method doesn't restrain a diagonalization target matrix to be combined into a hermitian symmetric matrix or a positive definite hermitian matrix and is wide in application; an alternative least square iterative algorithm based on combined diagonalization least square cost functions is adopted, and the structural characteristics of the target matrix system formed by the real-value target matrix are fully used to realize the combined diagonalization of a new target matrix system; The cost functions are solved by the alternative least square iterative algorithm, the estimate values of a mixed matrix are obtained, the blind source separation is realized, and the simulation results verify that the method provided is high in convergence precision.

The invention provides a method for realizing accurate identifying of a compound and screening of a differential component through automatic analysis of a complicated sample GC-MS, and belongs to thegas chromatography-mass spectrum combined data analysis. The method comprises the following steps: firstly, performing automatic extraction on a chromatographic peak under TIC and EIC, then determining an analysis range of each TIC chromatographic peak, and looking up EIC chromatographic peak information within the analysis range; clustering according to the EIC chromatographic peak information, and acquiring a representative chromatography contour spectrogram of each cluster; after screening, constructing an initial chromatography spectrogram matrix, performing optimized analysis on the initial chromatography spectrogram matrix by using a corrected multivariate curve resolution-alternating least square method, and acquiring the chemical components under each TIC chromatographic peak; andimporting a mass spectrum spectrogram of each chemical component obtained by analysis into a mass spectrum library for automatically matching the compound, and completing compound intelligent accurateidentification of a single sample GC-MS. The method has a good application value in the fields related to a GC-MS technology, such as scientific research, detection, and industrial application.

The invention discloses a signal two-dimensional DOA and frequency joint estimation method applied to an L-type array. The method comprises the following steps of (1) constructing a signal received bythe L-type array to a parallel factor trilinear model; (2) carrying out elementary row transformation on the model and carrying out block compression according to a compressed sensing theory; (3) using a trilinear alternating least square algorithm to carry out cyclic iterative decomposition on the compressed model until convergence so as to obtain the estimation of a parameter matrix; and (4) constructing the complete dictionary of frequency and angle estimations, and successively using a sparse recovery method to obtain the frequency and two-dimensional DOA estimations of a signal. The method has advantages that (1) the method is suitable for a uniform L array and a non-uniform L array, and the array is easy to realize during an actual application; (2) the calculated amount of an algorithm in the invention is small and simultaneously a capacity requirement to data storage is low; and (3) in the algorithm of the invention, by means of the parallel factor model, paired angle and frequency estimations can be acquired, and extra pairing does not need to be performed.

The invention discloses a double-base MIMO radar angle estimation method in colored noise background. Reception array data is matched and filtered, a cross covariance matrix of the reception data is estimated, and a four-order tensor model of the cross covariance matrix is established; a reception direction matrix, an emission direction matrix and a target covariance matrix are initialized, an ESPRIT algorithm is used to initialize the reception direction matrix and the emission direction matrix, and the target cross covariance matrix is initialized into a unit matrix; an alternative least square method is used to carry out iterative computation on the reception direction matrix, emission direction matrix and target covariance matrix till a convergence condition is met; and the least square method is used to estimate target DOD and DOA. Colored noise in the spatial domain is inhibited by using non-relevant characteristic of different fast-beat output noise, all freedom-degree information of array elements is used, the estimation precision is high, and there is no aperture loss.

The invention discloses a method of researching a reaction mechanism by utilizing MCR-ALS (Multivariate Curve Resolution-Alternating Least Squares) combined with an infrared online spectrum. The method comprises the steps of: inserting an infrared optical fiber probe in a reactor to collecting infrared spectrum data online, and performing wavelet transform denoising and baseline correction on the collected infrared spectrum data; then comparing subspaces obtained by separating different ingredient numbers through using two algorithms of main ingredient analysis and simplification self-mode; finally determining the ingredient numbers in a system to finish pretreatment; performing evolving factor analysis on obtained pretreatment data to obtain relative concentration discreet values of the ingredients in the system, substituting the relative concentration discreet values in an MCR-ALS program to perform alternating least-squares iteration for 200 times to obtain a concentration curve graph and a pure ingredient infrared spectrum of the ingredients of the finally optimized system, deducing the structure of each substance by combining an analyzed infrared spectrogram of each substance, and deducing the reaction mechanism by combining the concentration curve graph. According to the method, no additional intermediate capture agent is needed, thereby saving reagents; the online measurement mode guarantees accurate data and prevents information from being lagged; and only infrared online data are necessary, so that the measurement mode is simple, and various reactions can be measured.

The invention discloses a glyphosate concentration detection method based on Raman signals of chlorella pyrenoidosa. The glyphosate concentration detection method comprises the following steps: (1) acquiring Raman spectrum information of chlorella pyrenoidosa samples from water bodies with glyphosate of different concentration, and preprocessing the Raman spectrum information; (2) establishing a prediction model by virtue of alternating least square according to the concentration of glyphosate in the water bodies and the preprocessed Raman spectrum information; and (3) acquiring Raman spectrum information of a to-be-measured chlorella pyrenoidosa sample, processing the Raman spectrum information of the to-be-measured chlorella pyrenoidosa sample by virtue of a multivariate curve resolution method, and inputting the processed Raman spectrum information into the prediction model, thereby obtaining the concentration of glyphosate in a glyphosate water body from which the to-be-detected chlorella pyrenoidosa sample is collected. According to the detection method, the water sample does not need to be subjected to complicated preprocessing and chemical analysis, so that the operation steps are greatly simplified, and the detection time is shortened.

The invention discloses a tensor decomposition-based channel state information positioning fingerprint construction method. The method comprises the following steps of: firstly, expressing acquired channel state information (CSI) data as a three-dimensional image; regarding the three-dimensional image as a third-order tensor; then, combining a tensor decomposition algorithm based on a Parallet Factor (PARAFAC) analysis model and an ALS (Alternate Least Squares) iterative algorithm for noise reduction processing of the tensor; then, carrying out single-layer tensor wavelet decomposition on three dimensions of the CSI image by using a tensor wavelet decomposition algorithm, and calculating wavelet coefficients of wavelet sub-components by using an angular second moment; and finally, obtaining the CSI positioning fingerprint corresponding to each reference point coordinate. According to the method, the characteristic that high-order tensors can describe data information and structures isfully utilized, complex data is expressed in a tensor form, noise reduction and feature extraction of tensor images are finally achieved, and the data processing and analyzing capacity is improved.

The invention discloses a matrix and tensor combined decomposition-based car recommendation method and system. The method comprises the following steps of constructing a car scoring tensor, constructing a car manufacturer and supplier relationship matrix, and constructing a car product structure tree, wherein the car manufacturer and supplier relationship matrix and the car product structure are both complete and used for assisting in predicting a missing specific value in the tensor; according to the car product structure tree, introducing a tree-guided group lasso model for performing stipulation on a final loss function to obtain a weight; establishing a loss function, and performing iteration on the loss function by using an alternating least square method; performing derivation and zero setting on the loss function, and then calculating an iterative function of the matrix; recovering the tensor, namely complementing a missing value of the tensor; and for different users, based on elements in the complemented tensor X, recommending favorite car types of the users to the users according to a sequence of scores from high to low.

The invention discloses a bistatic MIMO radar angle estimation method under a joint error condition, comprising the following steps: performing matched filtering processing on data of a receiving array, and constructing one three-dimensional tensor model; then decomposing the three-dimensional tensor model by adopting a COMFAC algorithm and performing iterative computation on a receiving directionmatrix, a transmitting direction matrix and a target covariance matrix by utilizing an alternating least square method until convergence conditions are met, thus estimated values of the receiving direction matrix, the transmitting direction matrix and the target covariance matrix are obtained; and finally estimating DOD and DOA of a target by utilizing a least square method by virtue of a guidance vector of an auxiliary array element. The estimation method disclosed by the invention utilizes an array measurement multidimensional structure, so that precision of the estimation method is obviously better than that of a Reduced-MUSIC method; meanwhile, computational complexity of the estimation method is far below that of the Reduced-MUSIC method, and efficiency of the estimation method is higher than that of the Reduced-MUSIC method.

The invention discloses a method for fast measuring essence mixture ratio in cigarettes, which comprises the steps of: adopting all brands of sample essence for tabacco as a standard sample which is diluted with a solvent without ultraviolet absorption; diluting a sample essence for tabacco to be measured with the solvent without ultraviolet absorption; measuring the ultraviolet absorption spectrum of the standard sample and the sample essence for tabacco to be measured; taking the absorbance values corresponding to all wave lengths of the obtained ultraviolet absorption spectrum as independent variable matrixes, and taking the mixture ratios of all brands of essence for tabacco as dependent variable matrixes so as to construct an operational matrix; and operating the matrix by multivariate curve resolution-alternating least square method so as to obtain the actual mixture ratios of all brands of essence for tabacco. The method overcomes the defects of long analysis time, complex process, complicated detecting steps and incapability of on-site rapid test, and provides a simple method which reflects the actual mixture ratios of essence for tabacco, uses few testing instruments and can realize on-site rapid test, thus being capable of being widely used in tobacco manufacturing.