97results about How to "Good clustering" patented technology

A method for detecting faulty operation of a multi-variable system is described. The method includes receiving operational data from a plurality of components of the multi-variable system and processing the operational data in accordance with a plurality of dynamic machine learning fault detection models to generate a plurality of fault detection results. Each fault detection model uses a plurality of variables to model one or more components of the multi-variable system and is adapted to detect normal or faulty operation of an associated component or set of components of the multi-variable system. The plurality of fault detection results are output.

This disclosure generally provides molecules that specifically engage glucocorticoid-induced TNFR-related protein (GITR), a member of the TNF receptor superfamily (TNFRSF). More specifically, the disclosure relates to multivalent and / or multispecific molecules that bind at least GITR.

The invention relates to a high sliding window data stream anomaly detection method based on layered clustering, and aims to solve the problem that the accuracy of a data stream anomaly detection result is reduced due to influences of stale data and historical data. According to the method, by means of a layered clustering algorithm, the final clustering result cannot be considered during clustering, arrival data are processed at a higher speed, and a data volume of an off-line layer is greatly smaller than the number of original data due to the fact that the off-line layer only utilizes a clustering structure to respond to a user query result, so that the data can be effectively stored, and a more accurate clustering result can be obtained. As for a sliding window model, a clustering characteristic index histogram structure is adopted, so that insertion of new data and deletion of stale data can be better finished. A cosine coefficient is taken as a metric function, so that good clustering and anomaly detection results can be obtained. The high sliding window data stream anomaly detection method is applicable to fields of sensors, network click stream, share dealing and the like.

Exemplary embodiments of the present invention are directed to methods and program products for optimizing clustering of a design structure matrix. An embodiment of the present invention includes the steps of using a genetic operator to achieve an optimal clustering of a design structure matrix model. Other exemplary embodiments of the invention leverage the optimal clustering by applying a genetic operator on a module-specific basis.

The invention discloses a self-adaptive pre-segmentation method based on a gray scale and a gradient of an image and a gray scale statistic histogram. A threshold value sequence for image segmentation is self-adaptively generated according to the gray scale and gradient information of an image and the self gray scale statistic histogram information of the image, the continuous gray scale image is converted into a discrete gray scale image to realize the pre-segmentation of the image so as to facilitate the following image segmentation. The invention has less constraint condition, wide application range, low algorithm complexity, simple realization and high calculating speed.

User activity streams are used to automatically learn associations between logical objects and form logical groups. Search results are sorted based upon their relevance to current user activity. Combined with a graphical user interface component and object database, the invention automatically retrieves and display groups and objects related to the active object.

A method for detecting faulty operation of a multi-variable system is described. The method includes receiving operational data from a plurality of components of the multi-variable system and processing the operational data in accordance with a plurality of dynamic machine learning fault detection models to generate a plurality of fault detection results. Each fault detection model uses a plurality of variables to model one or more components of the multi-variable system and is adapted to detect normal or faulty operation of an associated component or set of components of the multi-variable system. The plurality of fault detection results are output.

This invention relates generally to molecules that specifically engage 41BB, a member of the TNF receptor superfamily (TNFRSF). More specifically, this invention relates to multivalent and multispecific molecules that bind at least 41BB.

For courses that deal with media content, such as sound, music, photographic images, hand sketches, video, conventional techniques for automatically evaluating and grading assignments are generally ill-suited to direct evaluation of coursework submitted in media-rich form. Likewise, for courses whose subject includes programming, signal processing or other functionally-expressed designs that operate on, or are used to produce media content, conventional techniques are also ill-suited. Instead, it has been discovered that media-rich, indeed even expressive, content can be accommodated as, or as derivatives of, submissions using feature extraction and machine learning techniques. In this way, e.g., in on-line course offerings, even large numbers of students and student submissions may be accommodated in a scalable and uniform grading or scoring scheme. Likewise, large collections of coursework submissions (whether or not graded or scored) or media content more generally, may be efficiently browsed and grouped using techniques described herein.

The invention belongs to the technical field of text semanteme processing, specially refer to a method and a system of relation characterizing, clustering and identifying based on semanteme of semantic space mapping. For the objects of the relationship to be extracted the invention comprises that grammar depending analysis is carried out for the sentence including two objects at first; then the analysis results is regarded as Graph, and the shortest path between two knots related two objects in Graph is calculated, to extract the relation of objects; afterwards, the words in the path is projected to semantic space, to accumulate and to obtain the vector expression of the relation in the semantic space; in the situation of multigroup of object couples, the relations are clustered by clustering method, to structure relatio model; according to semanteme vector expressing relation of input object couples and the semantic similar degree among relation models identification of relation is realized. The invention overcomes the shortcomings of the traditional method,such the sensitive factors as words deformation, synonym change, grammatical form changes etc. The accuracy and processing flexibility of identifying relationship is improved.

The invention relates to a hyperspectral small sample classification method based on a lightweight network and semi-supervised clustering. A lightweight network model is constructed by using a Point-wise convolution kernel, a Depth-wise convolution kernel and double loss. The Point-wise convolution kernel and the Depth-wise convolution kernel can greatly reduce the number of parameters, and reducethe demand for training samples in the network training process. The depth feature space can be more separable through the double-loss strategy, and classification and clustering in the depth featurespace are better facilitated. In addition, the semi-supervised approximate order clustering algorithm can select more self-confident pseudo tags, and more favorable conditions are provided for improving the network training effect. According to the method, autonomous extraction and high-precision classification of hyperspectral image depth features and label data are realized under the conditionof small samples.

The invention discloses an optimal splitting section search method based on semi-supervised spectral clustering. The optimal splitting section search method disclosed by the invention comprises the following steps of: constructing a detailed splitting section search model by virtue of related constraints of the target function of the minimum compound power flow impact, unit homology / separation, and the like at first, and then mapping the optimal section search optimization solution process to the relaxation solution obtaining process of constraint spectral clustering for static map segmentation, and finally selecting the optimal active splitting section through an improved PAM clustering algorithm. According to the process above, time complexity is reduced on the premise of no loss of whole network information, and the search result is accurate and effective.

The invention belongs to a method for identifying hematitization, and particularly discloses a method for identifying hematitization based on hyperspectral data. The method for identifying hematitization based on the hyperspectral data comprises the following steps that (1) hyperspectral image data are obtained and preprocessed; (2) a feature wave band of the hyperspectral image data is selected; (3) an image end member of the feature wave band of the hyperspectral data is extracted; (4) a spectral feature identification rule is built to distinguish and identify a hematitization end member and a ferritization end member; (5) map plotting is carried out on the hematitization end member and the ferritization end member through a mixed coordinate matching filtering method. By means of the method for identifying hematitization based on the hyperspectral data, hematitization and ferritization can be identified, the identification accuracy is high, and the detection limit of minerals is low.

A system and method for offloading network processes from main processors of a storage system and performing them on parallel processing modules. Embodiments of the present invention improve performance of a clustered storage system by performing certain network processes in an accelerator module of a storage system node. The accelerator module receives multi-protocol protocol data units (PDUs) from a network interface, performs protocol operations on the PDUs to form file system requests and passes the file system requests to a local D-module. If a file system request is directed to a non-local D-module in the cluster, the accelerator module repackages the request for transmission to the appropriate D-module and passes it back to the network without using local D-module processing resources or passing data over the system bus.

The invention discloses a local density spectral clustering similarity measurement algorithm based on Self-tuning. Through analysis of a similarity measurement method, a local density measurement method based on data neighborhood is provided. The method can be adopted to adaptively measure the scale of data and deal with data set clustering problems with a complex structure. The method has a good clustering effect compared with traditional spectral clustering methods and Self-tuning methods.

The present invention discloses a graph clustering method based on attribute fusion. A multi-layer attribute fusion model is constructed, the attribute characteristics and the structure relation of nodes in a graph are divided into different hierarchies, the data structure relation and the attribute characteristics are uniformly combined to the same bottom layer network for clustering operation, the data structure and attributes are subjected to weight fusion according to the node ballot mechanism in the clustering, and the weight coefficients of an attribute layer are subjected to adaptive changing to allow a final clustering result to reflect the original distribution of data so as to solve the influence problem of the clustering result of the setting of the initial value of the attribute layer and allow the final clustering to reach a better effect.

An object of the present invention is to enable optimal clustering for many types of noise data and to improve the accuracy of estimation of a speech model sequence of input speech. Noise is added to speech in accordance with noise-to-signal ration conditions to generated noise-added speech (step S1), the mean value of speech cepstral is subtracted from the generated, noise-added speech (step S2), a Gaussian distribution model of each piece of noise-added speech is created (step S3), the likelihoods of the pieces of noise-added speech are calculated to generate a likelihood matrix (step S4) to obtain a clustering result. An optimum model is selected (step S7) and linear transformation is performed to provide a maximized likelihood (step S8). Because noise-added speech is consistently used both in clustering and model learning, clustering for many types of noise data and an accurate estimation of a speech model sequence can be achieved.

The invention relates to a machine learning method for locally missing multi-view clustering based on matrix-guided regularization.The method fuses filling and clustering, fills missing kernel under the guidance of clustering, clusters with filled kernel, and introduces matrix-guided regularization when filling missing kernel. The method comprises the following steps: 1) obtaining target data samples and clustering target numbers, mapping the target data samples to multi-kernel space; 2) introducing matrix-guided regularization to establish regularized locally missing multi-kernel k-means clustering optimization objective function; 3) solving the regularized locally missing multi-kernel k-means clustering optimization objective function in a cyclic manner to realize clustering. Compared with the prior art, the present invention has the advantages of good clustering effect, low calculation amount and the like.

The invention discloses an image division method based on an inter-class maximized PCM (Pulse Code Modulation) clustering algorithm. The method comprises the following steps: carrying out classified labeling on pixel points of an input image according to a gray value; obtaining a clustering label when a clustering analysis method is used for dividing a target image; and carrying out performance evaluation on a label obtained by image division and an original label according to an evaluation index by a clustering method. The novel inter-class maximized PCM clustering algorithm considers the inter-class penalty, and parameters are adjusted and adjusted to enlarge the distance between class centers, so that the optimal classification of the pixel points in the image is realized.

The method utilizing a graph-like structure is comprises the following iteration:calculation of the potential of node nm, merging of two nodes ni and nj, as a function of the distances between the attributes of the key images and as a function of the temporal distance of these key images,calculation of the potential of each edge connecting the merged node to another node of the graph previously connected to nodes ni or nj,merging of the two nodes and validation of the new graph if the energy of this graph is less than the energy of the graph before merging.

The invention discloses an automatic coring and stringing device for hawthorns. The device comprises a rack, sugar-coated hawthorn sticks and a controller, the rack is provided with a coring station,and the rack is provided with a conveying mechanism used for conveying the sugar-coated hawthorn sticks to the position below the coring station one by one, a rotary disc mechanism used for conveyinghawthorns to the coring station one by one, and a coring mechanism used for coring the hawthorns located at the coring station. The rotary disc mechanism comprises an electric rotary disc and a rotating disc, a limiting mechanism is mounted on the lower end surface of the rotating disc, the coring mechanism comprises a first servo cylinder, a lifting plate, a punch and a push block, and a special-shaped plate is mounted at the lower end of the lifting plate. The coring and stringing device for hawthorns can replace manual work to core and string the hawthorns, so that the labor investment is reduced, the production efficiency is improved, and the industrial production of the sugar-coated hawthorns is facilitated; the device can realize an automatic coring function, is safe and sanitary, avoids the occurrence of an event that hands are stabbed by manual coring, reduces the breakage rate of hawthorn fruits, and ensures the quality of the hawthorn fruits.

An algorithm for modulating printer half-tone screening based on the tonal values of the region being reproduced is provided. The algorithm determines the tonal values for a particular region and assigns real number constants to that region. As the clustering screen is applied, the threshold values in the clustering screen are reduced (divided) by the real number constants thereby modulating the effect of the application of the clustering screen. For deep tonal regions, the clustering screen is applied at full strength, while in light tonal regions no clustering effect is applied. The algorithm results in a an image that has a greater dynamic range without the appearance of stair stepping while eliminating clustering artifacts from the lighter regions.

The invention provides a terylene low elastic net yarn target and a manufacturing method thereof which relates to a terylene low elastic net yarn target knitted on a double neilsbed raschel tricot machine, and a corresponding manufacturing method thereof. The method includes steps as follows: using 1-6D terylene yarn as a felted yarn raw material and 100-500d terylene FDY as a bottom yarn for knitting on the double raschel tricot machine, thermal finalization; terylene dyeing or printing; treating and back spraying or coating skidproof glue, then drying. The terylene low elastic net yarn manufactured by the method is knitted directly without spinning that leads to little producing procedure, short producing period, more reasonable and effective producing procedure and low cost; the produced fabric has better filling power, elastic recovery and stand property, and comfortable touch sense; the terylene low elastic net yarn target weights 1000g / m(2) per centiare that can compare beautiful with other carpet 2000g / m(2). The carpet can better serve market requirement.

The invention discloses a multimode biological identifying device and a method thereof. The multimode biological identifying device comprises three basic modes, namely, an infrared spectrum biological identifying technology, a hand biological identifying technology and a fingerprint biological identifying technology, and is characterized in that: by combining the infrared spectrum biological identifying technology, the hand biological identifying technology and the fingerprint biological identifying technology, a multimode comprehensive biological identifying system device is formed. According to the invention, a biological characteristic acquiring system is made by combining a laser technology, an infrared spectrum technology and a data acquiring technology, and the device consists of a biological signal acquiring unit, a signal transmitting and processing unit, a microprocessor unit, a control unit, a display and alarm unit and an interface unit. Due to the adoption of the multimode biological identifying method and the device thereof, the defect of low identifying rate of the conventional single-mode identifying system is overcome, a novel technical support is provided for an attendance system, a gate control system and a safety supervision system, high identifying accuracy is achieved, and the counterfeiting property of the system is ensured.

An algorithm for modulating printer half-tone screening based on the tonal values of the region being reproduced is provided. The algorithm determines the tonal values for a particular region and assigns real number constants to that region. As the clustering screen is applied, the threshold values in the clustering screen are reduced (divided) by the real number constants thereby modulating the effect of the application of the clustering screen. For deep tonal regions, the clustering screen is applied at full strength, while in light tonal regions no clustering effect is applied. The algorithm results in a an image that has a greater dynamic range without the appearance of stair stepping while eliminating clustering artifacts from the lighter regions.

A multi-level halftoning apparatus and method are provided. The multi-level halftoning apparatus includes a memory unit for storing a halftone table, and a comparison unit for determining a cell to be printed through comparison of an input tone with the stored halftone table. A sub-cell determined for printing after a reference main cell, which is a main cell in which a certain reference sub-cell is contained, becomes a sub-cell having the shortest distance from the reference sub-cell. The multi-level halftoning is processed based on distances between sub-cells constituting the halftone table so that cells determined for printing are optimally clustered, thereby improving the quality of printing.

The invention provides a streaming media content distribution method based on data feature dimension reduction coding, which comprises the following steps: step 1) extracting on-demand content data ofa server user, and performing historical content analysis to obtain a feature vector; 2) inputting the feature vector in the step 1) into a Word2Vec network, and training a content vector by using animproved Word2Vec model; wherein the improved Word2Vec model is obtained by training an original Word2Vec model and improving label data; 3) carrying out self-encoding training on the content vectorin the step 2) by utilizing a self-encoder model; 4) extracting an Encoder part in the auto-encoder model, and performing dimension reduction encoding on the content vector trained in the step 3) to generate a dimension reduction encoding content vector; and step 5) clustering the dimensionality reduction coding content vectors in the step 4) by using a k-means model, dividing the streaming mediacontent into n categories according to a clustering result, and respectively distributing the n categories to n edge servers of the corresponding content distribution network.

The invention provides a laser point cloud lane line extraction method and electronic equipment. The method comprises the steps of acquiring laser point cloud comprising a lane; based on the elevationvalue of each point in the laser point cloud, removing non-ground points from the laser point cloud to obtain ground point cloud; converting the ground point cloud into a binary image based on a grayvalue; performing region growing clustering on the binary image to generate at least one clustering region; performing principal component analysis on each clustering region to obtain a shape descriptor of each clustering region, and extracting ground points for generating lane lines from the ground points of each clustering region based on the shape descriptors; and fitting the ground points forgenerating the lane lines to generate the lane lines so as to manufacture a high-precision map. According to the method, the lane lines can be automatically and accurately extracted from massive point cloud data, the processing efficiency is high, and the extraction precision is high.

The invention relates to the field of data processing, and discloses an alarm data fusion method based on improved spectral clustering. The method comprises the following steps: preprocessing alarm data; grouping the alarm data according to attack types; calculating the similarity between every two alarms for the alarm data in each group by using an attribute similarity measurement method, and constructing a similarity matrix; clustering the alarm data by using a spectral clustering algorithm based on the similarity matrix to form clusters; performing threshold judgment on the alarms in the same cluster: if the threshold is reached, fusing the alarm data in the same cluster, and then inputting the fused data into a fused data set; if the threshold value is not reached, directly inputting the data into a fusion data set; and combining the fusion data sets of all clusters into a simplified alarm data set and outputting the simplified alarm data set. According to the method, better clustering fusion can be realized under the condition that the relation between alarms is not damaged, the information loss is reduced, and the false alarm rate of the alarm data can be reduced while the fusion rate is improved.

The invention discloses an image segmentation method of semi-supervised kernel k-mean clustering based on constraint pairs. The image segmentation method comprises the implementation steps: (1) selecting an image; (2) extracting texture features of the image; (3) generating a clustering object data matrix; (4) segmenting the clustering object data matrix; (5) initializing a clustering center; (6) calculating a distance; (7) judging whether the distance meets a constraint condition or not, if the distance meets the constraint condition, executing the step (8), and if not, executing the step (5); (8) calculating a mean; (9) judging whether the mean meets a termination condition, if the mean meets the termination condition, executing the step (10), and if not, executing the step (6); (10) generating a segmented image. According to the image segmentation method of the semi-supervised kernel k-mean clustering based on the constraint pairs, the texture features of the image are extracted, the image segmentation method of the semi-supervised kernel k-mean clustering based on the constraint pairs is used for segmenting the texture features, the stability of image segmentation is improved, and the more accurate image segmentation result is obtained.