117 results about "Nonlinear dimensionality reduction" patented technology

An object recognition system performs a number of rounds of dimensionality reduction and consistency learning on visual content items such as videos and still images, resulting in a set of feature vectors that accurately predict the presence of a visual object represented by a given object name within an visual content item. The feature vectors are stored in association with the object name which they represent and with an indication of the number of rounds of dimensionality reduction and consistency learning that produced them. The feature vectors and the indication can be used for various purposes, such as quickly determining a visual content item containing a visual representation of a given object name.

The invention discloses a tensor hyperspectral image spectrum-space dimensionality reduction method based on a deep convolutional neural network. The method comprises steps of: in view that it may significantly increase the parameter space of the deep convolutional neural network to directly use high-band tensor data, performing dimensionality reduction on the waveband of a normalized hyperspectral image by introducing a maximum likelihood intrinsic dimensionality estimation algorithm and principal component analysis to obtain a low-band hyperspectral image; converting the low-band hyperspectral image into a tensor low-band hyperspectral image by means of a window field, and keeping the spectrum and space information of each pixel; and performing spectrum-space dimensionality reduction on the tensor low-band hyperspectral image by means of the deep convolutional neural network in order that a characteristic subjected to the dimensionality reduction includes spectrum information and space information. The tensor hyperspectral image spectrum-space dimensionality reduction method may acquire a high overall classification precision and Kappa coefficient by using the spectrum characteristic and space field characteristic of the hyperspectral data.

The invention requests to protect an identification method for a human facial expression based on two-step dimensionality reduction and parallel feature fusion. The adopted two-step dimensionality method comprises the following steps: firstly, respectively performing the first-time dimensionality reduction on two kinds of human facial expression features to be fused in the real number field by using a principal component analysis (PCA) method, then performing the parallel feature fusion on the features subjected to dimensionality reduction in a unitary space, secondly, providing a hybrid discriminant analysis (HDA) method based on the unitary space as a feature dimensionality reduction method of the unitary space, respectively extracting two kinds of features of a local binary pattern (LBP) and a Gabor wavelet, combining dimensionality reduction frameworks in two steps, and finally, classifying and training by adopting a support vector machine (SVM). According to the method, the dimensions of the parallel fusion features can be effectively reduced; besides, the identification for six kinds of human facial expressions is realized and the identification rate is effectively improved; the defects existing in the identification method for serial feature fusion and single feature expression can be avoided; the method can be widely applied to the fields of mode identification such as safe video monitoring of public places, safe driving monitoring of vehicles, psychological study and medical monitoring.

Provided is a process of inferring insurability scores, the process including: receiving a request for an insurance comparison webpage; sending instructions to present one or more webpages of a web site having a plurality of user inputs configured to receive a plurality of attributes of the user; receiving the attributes; determining an insurability score with an insurability model based on the received attributes of the user; and sending instructions to display a value indicative of the insurability score.

The invention discloses a local spline embedding-based orthogonal semi-monitoring subspace image classification method. The method comprises the following steps of: 1) selecting n samples serving as training sets and the balance serving as testing sets from image data sets, wherein the training sets comprise marked data and unmarked data; 2) building an extra-class divergence matrix and an intra-class divergence matrix by using the marked data; (3) training data characteristic space distribution by using a whole and building a Laplacian matrix in a local spline embedding mode; 4) according to a local spline, embedding an orthogonal semi-monitoring subspace model, and searching a projection matrix to perform dimensionality reduction on the original high dimension characteristic; 5) building a classifier for the training samples after the dimensionality reduction by using a support vector machine; and 6) performing the dimensionality reduction on the testing sets by using the projection matrix and classifying the testing sets after the dimensionality reduction by using the classifier. In the method, the information, such as image sample marking, characteristic space distribution and the like, is fully utilized; potential semantic relevance among image data can be found out; and image semantics can be analyzed and expressed better.

The invention discloses a facial recognition method based on supervised neighborhood preserving embedding (SNPE) and a support vector machine (SVM). The method comprises a training process and a test process, and includes specifically the following procedures: a, a weight matrix for a given data sample group is constructed; b, according to the weight matrix obtained in the procedure a, a generalized characteristic vector problem about the data samples is solved and an embedding matrix projecting the data samples into a low-dimensional data space is searched; c, characteristic extraction is performed on the data samples with the embedding matrix to acquire characteristic data of the low-dimensional space; d, mode classification of the acquired characteristic data in the procedure c is conducted with the SVM to realize the type recognition of the data samples. The facial recognition method provided by the invention can solve the problems in the prior that the linear dimensionality reduction method cannot maintain well inter- and intra-type sample structures; the non-linear dimensionality reduction method has large computational amount; and the common classifier has over-learning and under-learning problems.

The invention discloses a hyperspectral image classification method based on image regular low-rank expression dimensionality reduction. The method includes the steps that a mean shift technology is used for conducting pre-segmentation on a hyperspectral image first, image regular low-rank coefficient expression is conduced on the hyperspectral image after pre-segmentation to obtain an image regular low-rank coefficient matrix, a characteristic value equation is constructed, a mapping matrix of the dimensionality reduction is studied, and original high dimensional data are transformed to low-dimensional space to be further classified. According to the hyperspectral image classification method, a hyperspectral image local manifold structure is excavated, the spatial distribution character of an original image is kept, effective dimensionality reduction space is studied, the classification accuracy of hyperspectral images is improved, computation complexity is lowered, the problems that the dimensionality of the hyperspectral image is too high so that the calculation amount can be large, and an existing method is low in classification accuracy are mainly solved, and the hyperspectral image classification method can be used for important fields such as precision agriculture, object identification and environment monitor.

The invention discloses a surface electromyogram signal identification method based on an LDA algorithm. The surface electromyogram signal identification method is used for identifying up to eight kinds of grabbing gestures. According to the surface electromyogram signal identification method, only two electromyogram electrodes are utilized to collect surface electromyogram signals of corresponding gestures from related muscles of a forearm of a tester at first, then original electromyogram signals are segmented in an overlapping windowing mode, and absolute mean values, variances and 4-order AR coefficients are extracted from various windows to serve as original electromyogram characteristics; the LDA algorithm is utilized to carry out dimensionality reduction on the original electromyogram characteristics, redundant information is removed to the maximum degree, useful information is kept, and characteristics after the dimensionality reduction are obtained; the mean value of dimensionality reduction characteristics of front and back adjacent windows is computed and is inputted to an LDA classifier, and effective identification for the eight kinds of grabbing gestures is achieved. According to the surface electromyogram signal identification method, the electromyogram signal identification rate for the various kinds of gestures is high, the whole signal processing process is simple in computation and low in time consumption, and the requirement for the real-time performance of an electromyogram control system is met.

The present invention is a method and an apparatus for reward-based learning of policies for managing or controlling a system or plant. In one embodiment, a method for reward-based learning includes receiving a set of one or more exemplars, where at least two of the exemplars comprise a (state, action) pair for a system, and at least one of the exemplars includes an immediate reward responsive to a (state, action) pair. A distance measure between pairs of exemplars is used to compute a Non-Linear Dimensionality Reduction (NLDR) mapping of (state, action) pairs into a lower-dimensional representation. The mapping is then applied to the set of exemplars, and reward-based learning is applied to the transformed exemplars to obtain a management policy.

The invention relates to a data mining improved type K mean value clustering method based on linear discriminant analysis, namely an LKM algorithm. Firstly, an LDA is adopted to conduct linear dimensionality reduction on an original n-dimensional data set A to obtain a one-dimensional data set Y, then a k mean value clustering algorithm is adopted to conduct clustering analysis on the data set Y after dimensionality reduction, and final results are output. The method that data dimensionality reduction is combined with the K mean value clustering method is adopted, and defects of the k mean value clustering algorithm on high-dimensional data are overcome through the data dimension reduction technology. The aims of lightening dimensionality curses and eliminating other uncorrelated attributes in high-dimensional space are achieved through data dimension reduction. Meanwhile, the performance of the k mean value clustering algorithm for processing the high-dimensional data is also improved, and the correlated defects of the k mean value clustering algorithm are overcome.

The invention discloses a behavior recognition method based on an action subspace and weight behavior recognition model. The behavior recognition method includes the following steps: A1, inputting a behavior video sequence to be tested, using a dynamic background detection method or a static background detection method for detecting background images, and using a background reduction method for obtaining foreground images; A2, extracting an outline of a moving object, and expressing outline characteristics correspondingly; A3, using a kernel principal component analysis (KPCA) algorithm for achieving nonlinear dimensionality reduction on high-dimensional characteristics in kernel guide subspace, and mapping behavior track in a low-dimensional space; and A4, using the behavior recognition model namely a world cancer research fund (WCRF) module for performing behavior recognition. An experimental result shows that a provided framework can recognize human behavior of personnel changes inside and outside an area along the time accurately and is strong in robustness of noise and other influencing factors.

The present application discloses a new technique which reduces the dimensionality of a feature model by re-use of data that has been obtained by a prior measurement. The data re-used from the prior measurement may range from parameters, such as geometrical dimensions, to more complex data that describe the electromagnetic scattering function of an underlying layer (for example, a local solution of the electric field properties).

The invention discloses an isolated digit speech recognition classification system and method combining a principal component analysis (PCA) with a restricted Boltzmann machine (RBM). First of all, a Mel frequency cepstrum coefficient (MFCC) is employed for combination with a one-order difference MFCC, and a voice dynamic characteristic of an isolated digit is preliminarily drawn off; then, linear dimension reduction processing is carried out on an MFCC combination characteristic by use of the PCA, and dimensions of a newly obtained characteristic are unified; accordingly, nonlinear dimension reduction processing is performed on the obtained new characteristic by use of the RBM; and finally, finishing recognition classification on a digit voice characteristic after nonlinear dimension reduction by use of a Softmax classifier. According to the invention, PCA linear dimension reduction, unification of the dimensions of the characteristic and RBM nonlinear dimension reduction are combined together, such that the characteristic representation and classification capabilities of a model are greatly improved, the isolated digit voice recognition correct rate is improved, and an efficient solution is provided for high-accuracy recognition of isolated digit voice.

The invention discloses a user driving behavior analysis method and system. The method comprises the steps of acquiring vehicle driving data of a plurality of users, and calculating a plurality of index items used for describing the driving behaviors of the users according to the vehicle driving data; Obtaining traffic accident information of a user; Taking the index items as characteristic variables, analyzing the correlation between each characteristic variable and the traffic accident information, and screening out N characteristic variables with the highest correlation with the traffic accident information to form an N-dimensional vector; Dimensionality reduction is carried out by using a nonlinear dimensionality reduction algorithm to obtain a variable set; Training a user driving behavior evaluation model by taking the variable set as an independent variable and the traffic accident information as a dependent variable; And evaluating the driving behavior of the to-be-analyzed user by using the user driving behavior evaluation model. The traffic accident information is used as a quantitative index for evaluating the driving behavior of the user, the user driving behavior evaluation model is trained, and the prediction and evaluation accuracy of the driving behavior of the user is improved.

The invention discloses a facial image dimensionality reduction classification method based on two-dimensional principal component analysis. The method comprises the steps of: firstly, performing image transformation on training samples; secondly, working out a covariance matrix according to transformed image data by using a 2DPCA (Data Processing Control Area) method, and working out the best projection matrix of a population covariance matrix; thirdly, performing spacial dimensionality reduction on projection of the training samples on the best projection matrix; and finally, classifying the training samples according to a nearest rule in a low-dimensional space. The method disclosed by the invention is quick to calculate a feature vector, accurate in calculation and high in recognition rate.

The invention discloses a semi-supervised dimensionality reduction method for high dimensional data clustering. The method comprises the following steps: (1) constructing a sample characteristic matrix; (2) constructing a constraint matrix; (3) constructing an iterative equation set and an iterative output transition matrix; and (4) obtaining a sample characteristic matrix after an operation of dimensionality reduction is performed. In the invention, through adding part of known class information as a constraint in the process of decomposing the sample characteristic matrix, and using an idea of concept decomposition, a coefficient matrix obtained by decomposition is used as the low-dimensionality representation of a high-dimensionality sample characteristic matrix, and when the low-dimensionality matrix is applied to clustering analysis, the clustering analysis becomes simple and effective; meanwhile, in the invention, data subjected to dimensionality reduction has a good interpretability; and compared with a dimensionality reduction method in the prior art, by using the dimensionality reduction method disclosed by the invention, the discrimination capacity of the clustering analysis can be further improved.

The invention relates to a similarity propagation and popularity dimensionality reduction based mixed recommendation method. According to the similarity propagation and popularity dimensionality reduction based mixed recommendation method, sparse data are processed in two phases; firstly, neighbors of the sparse data are expanded due to constant iteration of similar matrixes of users, resources and Tags through a similarity propagation method and accordingly elements for zero are filled; then a score algorithm in a search engine is introduced to calculate the Tag popularity in consideration of the problem that original data is provided with meaningless rubbish Tags, the tags with the popularity smaller than a certain threshold value are deleted to simplify data to perform dimensionality reduction on the matrix; recommendation results are diversified and the sparsity and cold starting problem can be relieved to some extent due to the fact that the recommendation based on contents and the collaborative filtering recommendation are combined. The similarity propagation and popularity dimensionality reduction based mixed recommendation method has the advantages of solving the problem of data sparsity in the individual recommendation process and being high in recommendation result accuracy, high in accuracy and high in reliability.

The invention discloses a face identification method based on a fuzzy rule. The face identification method is a novel feature extraction method and belongs to the field of face identification. The method comprises the steps that a face image is divided; sub-image variance and information entropy are calculated as the output of a fuzzy controller, so as to acquire the fuzzy weight of a corresponding sub-image; and LBP extraction is carried out on the sub-image to acquire a histogram vector and PCA dimensionality reduction is carried out on a sub-image vector after LBP extraction. The fuzzy weight has a significant feature for the image identification rate. Through PCA dimensionality reduction, the computation time of an algorithm is saved, and the timeliness of the algorithm is improved.

The invention provides a gait data-based identity recognition method. The method comprises the following steps of: firstly extracting gait profile curves of training samples and a to-be-recognized sample; processing the gait profile curves by utilizing a line-by-line scanning method so as to obtain a high-dimensional gait feature matrix; carrying out dimensionality reduction on the high-dimensional gait feature matrix by utilizing an improved smooth auto-encoder; and finally judging which category of training samples is nearest to the to-be-recognized sample by utilizing a nearest neighbor algorithm. According to the identity recognition method provided by the invention, a new gait feature is adopted, and the improved smooth auto-encoder and a nearest neighbor theory are utilized to carry out feature dimensionality reduction and similarity judgement, so that structure information in two-dimensional gait images can be sufficiently utilized to describe gait differences between different persons, thereby improving the gait information-based identity recognition correctness.

The invention discloses a data reduction method based on a stack type noise reduction self-coding neural network, which is characterized in that a reduction model of the stack type noise reduction self-coding neural network is constructed by the following steps of: 1, taking the output of the previous DAE as the input of the next DAE so as to achieve the purpose of layer-by-layer coding; Step 2, representing an original input sample by using a formula shown in the specification, and representing the coding condition of the DAE of the i layer d by the formula shown in the specification,and obtaining the coding condition of the DAE of each layer; And carrying out greedy training and fine tuning layer by layer, and in the fine tuning process, adjusting a cross entropy function of an initial parameter through a BP algorithm to ensure the minimization of a reconstruction error. Improved method-stack type noise reduction self-coding neural network algorithm adopting noise reduction self-coding network The is used for carrying out dimensionality reduction on a sample feature set, so that the complexity of various models is reduced, the classification effect of a classifier in machine learning application is improved, the operation cost of various learning algorithms is reduced, and the feasibility and high efficiency of reduction of the method are verified.

The invention discloses a supervising neighborhood preserving embedding face recognition method and system and a face recognizer. The supervising neighborhood preserving embedding face recognition method comprises the steps that initial dimensionality reduction is conducted on a training sample set, and then a primary dimensionality reduction training sample set and a primary dimensionality reduction training sample matrix are obtained; the class information of each training point in the primary dimensionality reduction training sample set is marked through a class divergence matrix; secondary dimensionality reduction is conducted on the primary dimensionality reduction training sample matrix through a secondary projection matrix, and then a secondary dimensionality reduction training sample matrix and a secondary dimensionality reduction sample set are obtained; a test sample is established, dimensionality reduction is conducted on the test sample twice, and then a secondary dimensionality reduction test sample is obtained; a secondary dimensionality reduction training sample which is closest to the secondary dimensionality reduction test sample is extracted, and the class label of the secondary dimensionality reduction training sample is given to the secondary dimensionality reduction test sample. Compared with the dimensionality reduction method in the prior art, the supervising neighborhood preserving embedding face recognition method can achieve supervised learning, and a high recognition rate is obtained.

The invention discloses a data clustering method based on dimensionality reduction and sampling, comprising steps of performing dimensionality reduction processing on a data set through a piecewise mean value algorithm, constructing a random function, performing random sampling from a large-scale clustering data set to obtain a working set with a relatively small scale, performing k-means clustering on the working set to obtain a random sampling clustering result, and performing classification on residual samples through measuring a relation between the residual samples and the obtained sampling clustering result. The data clustering method based on dimensionality reduction and sampling adopts the dimensionality reduction and sampling to reduce the number and dimensionalities of data samples which participate iteration, greatly reduces the complexity of the k-means algorithm under the condition that a good clustering effect is maintained, and realizes high efficiency clustering for big-scale data.

The invention relates to a multimode system feature dimensionality reduction method. The method comprises the steps that M sets of sample feature vectors under different operation modes Fn of a multimode system are acquired; the sample feature vectors are successively normalized to acquire normalized sample feature vectors; a multimode sample normalized feature matrix is constructed; a local linear embedding algorithm is used to carry out nonlinear dimensionality reduction on the normalized feature matrix; the most similar N-1 dimensional features of the same mode are selected; a multidimensional scaling algorithm is used to carry out linear dimensionality reduction on the matrix in the step (4); and two-dimensional feature matrixes with the largest difference in modes of different classes are selected. According to the invention, feature extraction technologies of non-linear and linear manifold learning algorithms are integrated; through effective feature dimensionality reduction of high-dimensional data, the difficulty of multimode system mode recognition is reduced; linearity and nonlinearity structures of high-dimensional data are maximally preserved; and domain characteristics and distance similarity of high-dimensional data are maintained.

Methods, systems, and apparatuses, including computer programs encoded on computer-readable media, for tokenizing n-grams from a plurality of text units. A multi-dimensional array is created having a plurality of dimensions based upon the plurality of text units and the n-grams from the plurality of text units. The multi-dimensional array is normalized and the dimensionality of the multi-dimensional array is reduced. The reduced dimensionality multi-dimensional array is clustered to generate a plurality of clusters that each cluster includes one or more of the plurality of text units.

The invention discloses an optimal material classification and recognition method based on hyperspectral data information maximization. The optimal material classification and recognition method comprises the following steps: (1) selecting training data from acquired hyperspectral data; (2) successively carrying out zero-mean, energy-keeping dimensionality reduction and unit normalizing pretreatment on the training data; (3) estimating a dimension matrix of line dimensionality reduction according to pretreatment data; (4) carrying out information maximizing row dimensionality reduction characteristic matrix calculation according to line-by-line dimensionality reduction dimension arrays; (5) carrying out classifier training according to a row-by-row dimensionality reduction characteristic matrix; (6) selecting an optimal characteristic matrix and an optimal classifier according to a training result; and (7) carrying out material classification and recognition on to-be-classified hyperspectral data according to the optimal characteristic matrix and the optimal classifier. The method provided by the invention has the advantages that the reduction of the hyperspectral data can be performed from a high-order statistics angle, and thus the high classification efficiency is achieved; a new classifier is easy to expand and add, so that the classifier with the excellent property is convenient to generate, and the material classification and recognition are well performed.

The invention discloses an infrared target robustness recognition method in combination with multi-feature dimensionality reduction and transfer learning. The method comprises: firstly providing, in view of the incomprehensive information coverage of a traditional infrared human target feature extraction method extracting a single feature, heterogeneous features for extracting different types of targets, including the shape feature and the brightness feature of the target image, so as to fully excavate the features of an infrared human target; secondly, using a principal component analysis method to reduce the dimensionality of the fused heterogeneous feature; and finally, designing an effective infrared human target classifier based on transfer learning to maximize the generalization performance and target recognition accuracy. Compared with the traditional infrared human target recognition method, the method provided by the invention improves the infrared human target recognition performance in a complex background.