168 results about "High dimensionality" patented technology

Data mining techniques are provided which are effective and efficient for discovering useful information from an amorphous collection or data set of records. For example, the present invention provides for the mining of data, e.g., of several or many records, to discover interesting associations between entries of qualitative text, and covariances between data of quantitative numerical types, in records. Although not limited thereto, the invention has particular application and advantage when the data is of a type such as clinical, pharmacogenomic, forensic, police and financial records, which are characterized by many varied entries, since the problem is then said to be one of “high dimensionality” which has posed mathematical and technical difficulties for researchers. This is especially true when considering strong negative associations and negative covariance, i.e., between items of data which may so rarely come together that their concurrence is never seen in any record, yet the fact that this is not expected is of potential great interest.

A clustering method for high-dimensionality data includes identifying a set of nearest neighbors of a point in a multidimensional space and determining the centroid of the set of nearest neighbors, where the centroid is a member of the set of nearest neighbors. The method is then repeated using the neighbors identified around the computed centroid. In one embodiment, the method may terminate when the computed centroid becomes stationary over successive iterations. The resulting centroid may be returned as a mode of the data set. Points of the data set having common modes may be assigned to the same cluster.

A classifier training method and apparatus for training, a linear classifier trained by the method, and its use, are disclosed. In training the linear classifier, signatures for a set of training samples, such as images, in the form of multi-dimension vectors in a first multi-dimensional space, are converted to a second multi-dimension space, of the same or higher dimensionality than the first multi-dimension space, by applying a set of embedding functions, one for each dimension of the vector space. A linear classifier is trained in the second multi-dimension space. The linear classifier can approximate the accuracy of a non-linear classifier in the original space when predicting labels for new samples, but with lower computation cost in the learning phase.

The invention discloses an examiner identification system based on bionic and biometric identification, using synthetically various biometric identification methods based on high-dimension space geometric shape adaptive coverage theory to achieve identification of examiner identity. First of all, through an acquisition equipment, gripping-pen fingerprint is obtained, on-line signature and facial image, and then the data is mapped into high-dimension space observation point after feature extraction, last according to similar sample point continuity in the high-dimension space, through the relation between the observation point and sample set coverage area to obtain different biological characteristic network match degree, then through match degree fusion decision algorithm to identify identity of the examiner, and through automatic addition of new verification data to achieve sample set dynamic updating and trend forecast. Identification of the invention is fast, result is accurate, and the invention is not only suitable for examiner identification in existing examination mode, but also has broader application in the future machine- examination mode.

The invention discloses a method for automatically identifying a breast tumor area based on an ultrasound image. The method comprises the following steps of acquiring the ultrasound image of the breast, and preprocessing the ultrasound image; segmenting the ultrasound image subjected to preprocessing through an image segmentation method to obtain a plurality of segmented subareas; extracting a grey level histogram, texture features, gradient features and morphological features of the ultrasound image, and combining the grey level histogram, the texture features, the gradient features and the morphological features of the ultrasound image with two-dimensional position information to obtain high-dimensionality feature vectors; selecting the most effective feature subset of the high-dimensionality feature vectors through feature ordering based on biclustering and a selection method; performing learning classification on the selected most effective feature subset through a classifier, and then automatically identifying the breast tumor area. By means of the method, the breast tumor area can be identified automatically from segment results of the breast tumor ultrasound image, therefore, automation performance of computer-aided diagnosis is improved, manual operation of clinical doctors is reduced, and subjective influence of clinical doctors is reduced.

The invention relates to a pedestrian detection method based on video processing. The pedestrian detection method comprises the steps of (1) extracting a foreground image, extracting a moving object image of each frame of a video, marking the image and storing the image into a storage in sequence, using a background model to extract a background, enabling the model to adopt the gauss mixing model, (2) conducting preliminary screening on the foreground, selecting shape features of a pedestrian for conducting identification, (3) accurately identifying the foreground, selecting HOGs to conduct feature extraction on the foreground image after preliminary screening, then using a low dimensionality soft output SVM pedestrian classifier to conduct classification, and judging whether the pedestrian exists or not. The pedestrian detection method further comprises the step of (4) conducting error correction processing in a secondary thread. As for the foreground image with low dimensionality soft output SVM pedestrian classifier soft output results which are ambiguous in belonging classification, a high dimensionality SVM classifier is called in the secondary thread for recognition processing. The pedestrian detection method based on video processing improves the detection accuracy and is good in real-time performance.

The invention discloses a joint optimized scheduling method for multiple types of generating sets of a self-supply power plant of an iron and steel enterprise, and belongs to the technical field of energy optimized scheduling of the iron and steel enterprise. Influence of fuel types and gas mixed burning amount on energy consumption of the sets is taken into consideration in construction of a set energy consumption characteristic model, fitting is performed under different gas mixed burning, and the accuracy and representativeness of the model are improved; and influence of the fuel cost, time-of-use power price and surplus gas dynamic change on the generating cost is considered comprehensively in construction of an optimized scheduling model, meanwhile, various constraint conditions including power balance constraint, generating set self-running constraint, purchased power quantity constraint, gas supply constraint, variable load rate limit and the like are considered, and the performability of a generation schedule is guaranteed. Optimization solution is performed on the models by adopting the adaptive particle swarm optimization algorithm, the problems of high dimensionality, nonconvexity, nonlinearity and multiple constraints of the power generation scheduling of the self-supply power plant can be well solved, power production optimization and purchasing rationalization are realized, surplus gas is sufficiently used, and the power supply cost is reduced to the greatest extent.

The invention discloses a narrow-line-width high-dimensionality quantum entanglement light source generating device. The device comprises a pumping laser unit, a crystal unit, a filter unit and a collection and analysis unit, the pumping laser unit is used for generating pumping light needed in the process of spontaneous parametric down-conversion of the crystal unit, the crystal unit is used for receiving pumping light emitted by the pumping laser unit and utilizing the spontaneous parametric down-conversion process to generate a high-dimensionality quantum entanglement light source carrying orbital angular momentum information, the filter unit is used for filtering pumping light in the high-dimensionality quantum entanglement light source and narrowing line width of the same, and the collection and analysis unit is used for collecting the narrow-line-width high-dimensionality quantum entanglement light source after being processed by the filter unit and measuring entanglement characteristics of the same. The narrow-line-width high-dimensionality quantum entanglement light source generating device has the advantages of convenience in adjusting, high-dimensionality correlation and can be used in various application fields like quantum communication, quantum networks, quantum passwords and quantum physical testing.

The invention belongs to the technical fields of cloud computing and data mining, and particularly discloses a massive high-dimension data clustering method for a MapReduce platform. In the method, each dimension of raw data is split, and clustering is performed by utilizing small split non-null grids instead of points in the raw data so as to reduce a data scale. The clustering is realized by utilizing an open source of MapReduce, so that the whole clustering process can be finished in parallel on a distributed cluster, and the limitations of a single-machine algorithm to storage and computation are broken. In the clustering process, the thought of a K-mediods algorithm is adopted, and a highly-efficient Euclidean distance computation method is put forward. The method is applied to the processing of massive high-dimension data. A user can perform manual regulation on the algorithm according to the computational capability of the cluster, the expected time of the algorithm and requirements on clustering accuracy. The needs of different users are satisfied.

The invention relates to a chaotic neural network encryption communication circuit, in particular to an encryption communication circuit based on a chaotic neuron network with a time delay state. A clear text signal i(t) of a transmission end drives a first time delay chaotic neural network system through a reversed phase amplification circuit, the first time delay chaotic neural network system outputs a chaotic signal x(t), the chaotic signal x(t) and the clear text signal i(t) are overlapped to generate a signal x(t)+i(t), an encryption transmission signal s(t) is generated through an encryption scheme circuit; and the encryption transmission signal s(t) is transmitted to a receiving end through a transmission channel, the signal x(t)+i(t) is solved through corresponding decryption scheme circuit to drive a second time delay chaotic neural network system, the second time delay chaotic neural network system generates a corresponding chaotic signal y(t) synchronous with the chaotic signal x(t), and the signal x(t)+i(t) is subtracted from the chaotic signal y(t) to obtain a clear text signal r(t). The chaotic neural network encryption communication circuit overcomes the defects that confidentiality of a common low-dimensional chaotic system is poor and a high-dimensional chaotic system is difficult to be physically implemented, implements encryption transmission function of theclear text signal, and effectively simplifies a real circuit device.

A method receives the symbol transmitted over a channel, selects, from a constellation of codewords, a first codeword neighboring the received symbol and a set of second codewords neighboring the first codeword, and determines a relative likelihood of each second codeword being the transmitted symbol with respect to a likelihood of the first codeword being the transmitted symbol. Next, the method determines an approximation of a log-likelihood ratio (LLR) of each data bit in the received symbol as a log of a ratio of a sum of the relative likelihoods of at least some of the second codewords having the same value of the data bit to a sum of the relative likelihoods of at least some of the second codewords having different value of the data bit and decodes the received symbol using the LLR of each data bit.

The invention provides a multispectral calculation reconstruction method which comprises the steps of generating a two-way multispectral image and employing a photographing device toacquire the two-way multispectral image to obtainmultispectral information of sampling points; according to the multispectral information of the sampling points generating a corresponding multispectral information matrix of the sampling points and allowing the multispectral information matrix of the sampling points to be subjected to dictionary learning with sparse constraint to generate a spectral dictionary; and reconstructing the spectral information of non-sampling points in the two-way multispectral image under sparse prior constraint. The invention also provides a multispectral calculation reconstruction system which comprises an image acquisition device, a dictionary learning device and a spectral information reconstruction device. The method and the system provided by the invention employthe inherent law of the multispectral information, scene materials and the sparsity of light source spectrum, thus the reconstruction of the multispectral informationbeing simple and intuitive, the needed scene spectrum sampling points being less, and realizing high dimension multispectral data collection based on a compression perception theory.

A two-level text similarity calculation method based on subjective and objective semantics is characterized in that text is divided into a topic and a main body, a topic-word vector is built by filtering, a main body-word vector with low dimensionality is built by extracting keywords, a word semantic similarity calculation method achieving subjective and objective combination is used for calculating word vector similarity so as to obtain the topic similarity and the main body similarity respectively, and therefore the text similarity is obtained; the word semantic similarity is calculated on the basis of word-text indexes of HowNet and a corpus, so that words are expressed concisely, and calculation results accord with not only subjective concepts but also objective semantic environments; during calculation of the text similarity, equal importance is attached to the topic and the main body, the word semantic similarity calculation method achieving subjective and objective combination is used, a text-word vector with high dimensionality is avoided, text information is extracted fully, accuracy of text similarity results is improved, and the two-level text similarity calculation method is suitable for text similarity analysis under various circumstances.

A method is described for the automatic determination of a planogram in a vending machine using image processing. Novel steps include the use of a Gaussian map in a two-dimensional color space, such as the HS plane, to create high-dimensionally color vectors for all and for selected portions of images. Multiple feature detection / extraction algorithms are run between multiple idealized reference images for a product and one image from one vending machine coil location. The large resulting candidate feature list is pruned in a series of steps using both color and gray-scale color vectors and small area image matching around features. Remaining candidate features are ranked by a RANSAC outlier removal step, with the top ranked product then being the correct product in that coil in the planogram. Steps are repeated for all coils in a vending machine.

The invention provides a graph classification method based on graph set reconstruction and graph kernel dimensionality reduction. The method comprises the steps of: 1) performing frequent sub-graph mining on a graph data set used for training, and performing discriminative sub-graph screening on obtained frequent sub-graphs with the emerging frequentness differences of the sub-graphs in a positive class and a negative class; 2) reconstructing the original graph set with selected discriminative frequent sub-graphs; 3) obtaining a kernel matrix for describing the similarity between every two graphs in the newly-reconstructed graph set by using a Weisfeiler-Lehman shortest path kernel method, and based on class label information of training graphs, performing dimensionality reduction on high-dimensionality kernel matrixes by using a KFDA method; 4) training graph data projected to a low-dimensionality vector space based on an extreme learning machine to build a classifier; 5) standardizing graph data requiring classification, projecting the data to a low-dimensionality space obtained through training and inputting the projected data to the classifier to obtain a classification result. The method can directly classify graph data without class labels and guarantee high classification accuracy.

The invention discloses a planetary gear fault diagnosis method based on dual-tree complex wavelet transform-entropy feature fusion. The method comprises the following steps of collecting integration simulation experiment table data and acquiring a planetary gear shell original vibration signal; using dual-tree complex wavelet transform to decompose an original vibration signal and extracting a signal component of each frequency band; constructing an entropy feature extraction model from multiple angles and acquiring a high-dimension original feature; using a nucleus Fisher discriminant analysis method to carry out dimension reduction processing on an original feature set formed by a plurality of entropy features, determining a group of optimum discriminant vectors, extracting a projection of the original feature in the optimum discriminant vectors and taking as a sensitive fault feature so as to determine a fault type; verifying a necessity of describing feature information from the multiple angles and multiple spaces and validity of carrying out feature dimension reduction by using a KFDA method based on that. The method is suitable for the non-linear and non-stable planetary gear vibration signal with a high coupling feature. By using the method, the sensitive fault feature can be effectively extracted and accurate diagnosis of the planetary gear is realized.

One embodiment of the present invention provides a system for generating a classifier to detect patterns in a data sequence. During operation, the system receives the data sequence, which represents a sequence of measurements of a phenomenon. The system transforms the data sequence into a feature sequence that is of a higher dimensionality than a dimensionality of the data sequence, and the feature sequence is a sequence of feature vectors each created from contiguous members of the data sequence. Next, the system generates a graph where each node of the graph corresponds to a feature vector. The system converts the generated graph into a two-dimensional graph. Subsequently, the system displays, to a user, the two-dimensional graph. The system receives user input indicating that a region of the two-dimensional graph corresponds to a pattern associated with the feature sequence, and then generates a classifier based on the received user input.

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 discloses an information transferring model based fault tracing method for process industrial complex electromechanical system. On the basis of mass high-dimensional system operating state monitoring data, by adopting a data analysis method and taking an information transferring relation among different monitoring variables as a measure of a system information model coupling relation, a system fault process recognition method and a tracing method are provided in comprehensive consideration of influences of process industrial feedback control on a system fault tracing process. According to the method, a unique root event of a system fault can be traced from any abnormal information monitoring points, and the fault tracing process is independent of system physical topology priori knowledge. The processing process can be partially and directly applied to information modeling of the process industrial complex electromechanical system, tracing of system fault causes can be realized, the digital monitoring level of enterprises is raised, and scientific maintenance is facilitated.

Under control of composite sequence generated by plaintext, sub system of autoregressive non-linear digital filter modulates plaintext to chaos locus in high dimension in composite filter. Hashed value of plaintext is produced by quantizing chaos locus in coarse granulation. Iterative initial point of composite filter is as cipher key of algorithm, which satisfies requirement of security of Hash algorithm with cipher key. Sensitivity and traversing characteristic on initial value of chaos in high dimension makes hashed result sense to plaintext exceedingly. Moreover, hashed result is distributed in hashed space evenly. The composite sequence increases randomness selected by sub system of filter so as to guarantee complex sensitive nonlinear relation between iterative locus and initial condition. Thus, the invention possesses better scrambling, and stronger capability for anti deciphering. Features are: simple and fast algorithm, easy of modularized realization.

The invention provides a nuclear power device fault diagnosis method based on local linear embedding and a K-nearest neighbor classifier. The method comprises steps of (1) acquiring operation data of a nuclear power device in steady-state operation and typical accident states as training data; (2) using the mean-variance standardization method, carrying out dimensionless standardization processing on the training data to obtain high-dimension sample data; (3) using the local linear embedding algorithm, extracting low-dimension manifold structures of the high-dimension sample data so as to obtain low-dimension characteristic vectors; (4) inputting the low-dimension characteristic vectors into a K-nearest neighbor classifier to carry out classification training; (5), acquiring real-time operation data of the nuclear power device, and repeating the steps of (2) and (3); and (6) using the trained K-nearest neighbor classifier to make decisions for classification of the characteristic vectors. According to the invention, by taking advantages of the nonlinear manifold learning method in the aspects of characteristic dimension reduction extraction, the provided method is suitable for fault diagnosis of nonlinear data high-dimension systems, and has quite high fault diagnosis accuracy.

A speech verification process involves comparison of enrollment and test speech data and an improved method of comparing the data is disclosed, wherein segmented frames of speech are analyzed jointly, rather than independently. The enrollment and test speech are both subjected to a feature extraction process to derive fixed-length feature vectors, and the feature vectors are compared, using a linear discriminant analysis and having no dependence upon the order of the words spoken or the speaking rate. The discriminant analysis is made possible, despite a relatively high dimensionality of the feature vectors, by a mathematical procedure provided for finding an eigenvector to simultaneously diagonalize the between-speaker and between-channel covariances of the enrollment and test data.

The invention discloses a feature selection method and device of high-dimension data. The method comprises the following steps: obtaining an original data set to be processed, wherein the original data set comprises a feature set, a plurality of samples and a category set, and the category set comprises the category of each sample; calculating to obtain a MIC (Maximum Information Coefficient) between each feature in the feature set and the category set, and the redundant value of each feature and a selected feature subset; and according to the MIC and the redundant value, obtaining the effective value of each feature, and selecting the feature subset from the feature set according to the effective value. The MIC is introduced into feature selection, and the feature is effectively evaluated on the basis of the MIC so as to select features according to the effective value generated by evaluation. Compared with the prior art, the feature selection method can effectively improve accuracy for high-dimension data feature selection.

The invention relates to a Chinese author identification method based on a double-layer classification model and a device for realizing the Chinese author identification method, belonging to the field of information security. Aiming at the problem of low identification accuracy caused by excessive authors, an author grouping layer is added in an author identification model; each author is represented into an author vector; authors are grouped by a clustering algorithm; a second layer is an author identification layer; a dependence relationship, a function word, a punctuation mark and a word class mark are extracted from the second layer to use as characteristics; and author identification is carried out in the group. According to the method or the device, the problem that the identification accuracy is lowered because of excessive authors can be effectively solved. Meanwhile, with a proposed characteristic dimensionality reduction and optimization method based on a main ingredient analysis method, the problem that the identification accuracy is affected by noise comprised by a high-dimensionality characteristic vector is solved. The Chinese author identification method can be applied to the author textual research field of a literature and also can be applied to the field of information security, such as copyright protection.

The invention discloses a power generation scheduling method based on a high-dimension wind-electricity prediction error model and the dimensionality reduction technology. The method comprises steps of: acquiring history output data of each hour in one year of multiple wind power plant and corresponding point prediction data; using a mixed skewness model to carry out modeling on accumulated distribution functions of actual output and predicted output of each wind power plant; using the CDF of each wind power plant to convert the actual output value and the prediction value into data points distributed in 0-1 intervals; by matching all data points obtained in the previous step, finding out the optimal Copula function and carrying out parameter estimation; establishing high dimension condition probability model of multiple wind power plant prediction errors, and obtaining edge condition probability models subjected to dimensionality reduction trough edge conversion; and according to the edge condition probability models of the wind power plant prediction errors, calculating the current scheduling plan of the generator unit and the rotation standby capacity. Compared with the common gauss distribution and beta distribution, the power generation scheduling method is quite high in precision, and effects of relevance between multiple wind power plants can be considered.

The invention discloses an airspace operation situation evaluation and classification method based on fuzzy reasoning and belongs to the technical field of airspace situation evaluation and classification. The method comprises the following steps: 1, processing airspace operation situation samples in a to-be-processed sector; 2, establishing a primary fuzzy reasoning system based on the airspace operation situation samples in the to-be-processed sector; 3, optimizing interpretability and accuracy of the fuzzy reasoning system on the basis of a multi-target population adaptive immune algorithm.The method provided by the invention can aim at large-scale and high-dimensionality sector operation data and center on airspace operation situation evaluation accuracy and interpretability, the multi-target immune optimization algorithm is used, and the airspace situation evaluation accuracy is optimized. In addition, when the immune algorithm is realized, the condition that the fuzzy matrix scale is in exponential growth in the process of processing high-dimensionality data is avoided, time complexity and space complexity needed by the algorithm are greatly reduced, and the convergence precision is improved.

The invention provides a discriminant human motion identification method based on a Gaussian process latent variable model and a hidden condition random field. The method mainly comprises that skeletal structure and motion information of the human body are obtained via motion capturing technology or Kinect motion sensing technology when motion data is obtained; when motion characteristics are extracted, the Gaussian process latent variable model added with a dynamic process and sparse approximation is used to obtain structure of high-dimension motion information in low-dimension hidden space and further to represent the motion characteristic; and when the human motion is identified, the discriminant hidden condition random field is used to model characteristics of sequential motion data, and motions are classified. According to the invention, characteristics of the human motions can be visualized, information among the sequential motion data can be used effectively, the human motions can be identified in high precision, and the method is suitable for the field of real-time human motion identification.

A nearer neighbor matching and compression method and apparatus provide matching of data vectors to exemplar vectors. A data vector is compared to exemplar vectors contained within a subset of exemplar vectors, i.e., a set of possible exemplar vectors, to find a match. After a match is found, a probability function assigns a probability value based on the probability that a better matching exemplar vector exists. If the probability that a better match exists is greater than a predetermined probability value, the data vector is compared to an additional exemplar vector. If a match is not found, the data vector is added to the set of exemplar vectors. Data compression may be achieved in a hyperspectral image data vector set by replacing each observed data vector representing a respective spatial pixel by reference to a member of the exemplar set that “matches' the data vector. As such, each spatial pixel will be assigned to one of the exemplar vectors.

The invention relates to a high dimension data index structure design method based on a solid state hard disk, which includes the following steps: an index structure is divided into an original R tree area and a node difference log area which are respectively used for storing original version data and difference logs of the original version and the latest version; a Hashtable is designed in the internal storage to store nodes and update information in corresponding relation to the position of the node difference log area; once a new update is finished, update log of an earlier time at the node is read out, and then the log and the current log are combined and stored again to serve as all updated logs at the node so far. Based on the previous R tree, the node difference log area is added, node difference logs are designed, and random update operation is changed to random update so as to improve update efficiency. The node difference logs are capable of enabling logs aiming at a certainnode to be stored within a certain range, and read operation of the node difference log R tree is at most twice that of the previous R tree.

The invention discloses a rolling bearing variable-work-condition fault diagnosis method based on visual cognition, and relates to a rolling bearing variable-work-condition fault diagnosis technology. The method comprises the following steps of converting rolling bearing vibration signals under the variable work conditions into a two-dimensional image by using a recurrence plot technology; performing feature extraction on the two-dimensional image by utilizing an SURF (speed up robust features) algorithm to obtain the vision invariability high-dimension fault feature vector; performing dimension reduction processing on the high-dimension feature vector by using an equal-distance mapping Isomap algorithm to obtain the low-dimension stable feature vector; using an SVD (singular value decomposition) algorithm for extracting the feature matrix singular value built by the low-dimension stable feature vector to form the final feature vector; performing fault classification on the final feature vector by using the trained classifier; performing fault diagnosis on the rolling bearing under the variable work conditions. The invention provides a novel solution for the rolling bearing fault diagnosis.