34 results about "Kernel model" patented technology

An utterance is received from a user specifying a location attribute and a landmark. A set of candidate locations is identified based on the specified location attribute, and a confidence score can be determined for each candidate location. A set of landmarks is identified based on the specified landmark, and confidence scores can be determined for the landmarks. An associated kernel model is generated for each landmark. Each kernel model is centered at the location of the associated landmark on a map, and the amplitude of the kernel model can be based on landmark attributes, landmark confidence scores, characteristics of the user, and the like. The candidate locations are ranked based on the amplitudes of overlapping kernel models at the candidate locations, and can also be ranked based on confidence scores associated with the candidate locations. A candidate location is selected and presented to the user based on the candidate location ranking

Core model processing of a processor model PE1 and a processor model PE2 is serialized. Therefore, processing time for the inter-core-model communication is required between the core model processing of a first processor model and the core model processing of a second processor model. The inter-core-model communication processing is performed such that the inter-core-model communication required for the simulation processing of a multi-processor model is performed in parallel with the core model processing.

In various implementations of the invention, a model of an optical proximity correction process is employed to determine potential adjustments to a layout design for a mask that might resolve potential errors an image resulting from application of the mask in an optical lithographic process. In various implementations of the invention, corrected mask shapes, such as for example optical proximity corrected mask shapes, and associated printed image contours are generated through use of a model. Subsequently, the associated printed image contour and a desired printed image contour may be used to determine various edge segment adjustments to the corrected mask shapes that would realize the desired printed image contour. In various implementations of the present invention, the model for generation of the corrected mask shapes and the associated printed image contour is a square kernel model. With various implementations of the invention, the kernel represents a grey scale map wherein each pixel of the map is generated based on the desired displacement relative to the displacement to be modeled. For example by application of linear regression techniques. As a result, printed image contours and corrected mask shapes may be generated based upon an input layout design, wherein potential adjustments to the mask may be determined based upon a desired printed image contour.

The invention relates to the technical field of power system scheduling and operation, in particular to a micro-grid short-term load prediction method based on long-term and short-term memory and self-adaptive lifting, which comprises the following steps: step 1, calling historical load data; step 2, integrating the data to obtain a training set and a test set; step 3, performing integrated empirical mode decomposition and adjustment on the training set and the test set, and outputting a training sample set and a test sample set; 4, establishing a combined prediction kernel model, and settinghyper-parameter values; 5, inputting the training sample set, and outputting a prediction result; 6, setting the cycle index N, and entering the step 7 when the actual cycle index is greater than N; if the actual cycle index is smaller than N, entering the step 5; 7, calculating a root-mean-square error, judging whether the root-mean-square error is stable or not, entering the step 9 if the root-mean-square error is stable, and entering the step 8 if not ; 8, adjusting hyper-parameters, and entering the step 5; and step 9, inputting the test sample set, and outputting a prediction result. Themethod is high in prediction precision, small in error, high in adaptability and high in practicability.

The present invention discloses a multi-agent based route selection simulation system in a vehicle online environment. The system consists of two parts: foreground and background. A foreground interface is a human-computer interaction module, and a background program is responsible for implementing a function of the simulation system. The system comprises a route network generation and initialization model, a vehicle generation and initialization model, a simulation kernel model and an information acquisition and transfer model. A simulation kernel is core rules for driving a vehicle to run in a route network and comprises two layers of rules: micro rules and macro rules. The present invention studies overall design of the multi-agent simulation system for a route selection behavior of a traveler in the vehicle online environment, two layers of action rules of the vehicle-online multi-agent and information acquisition and transfer in the simulation system and the like; and starting from each vehicle action rule, the macro state of the system is reflected, so that simulation of the complicated dynamic system is feasible and effective.

The invention discloses a noise diagnosis method for an electrical equipment break-down arc. The method comprises the following steps that A, an arc sound signal of the electrical equipment is collected, and pre-emphasis processing, FIR digital filtering and framing are carried out on the arc sound signal; B, short-time energy and short-time zero-crossing rate of each frame of the arc sound signalare calculated, short-time average energy and short-time average zero-crossing rate are determined, and an arc sound signal abnormal interval is detected by adopting a dual threshold determinate method; C, filtering is carried out on the arc sound signal abnormal interval by adopting M FIR filters, an energy value of each frequency domain sub-band is calculated to form an M-dimensional feature vector; D, a primary linear kernel model is established by using sample set data and an optimal parameter of a kernel function, an optimal linear kernel model is obtained by modifying the parameter repeatedly, the M-dimensional feature vector is diagnosed by using the optimal linear kernel model to determine whether or not a running state of the electrical equipment is normal. The noise diagnosis method for the electrical equipment break-down arc has the advantages that on-line detection and real-time alarming of the running state of the electrical equipment are achieved, and the damage and probability of the break-down arc are reduced.

The invention discloses an M module low-level (LL) driver layer realization method for M module-based local area network (LAN)-based extensions for instrumentation (LXI) equipment, which belongs to the field of automatic test and aims to solve the problem of independence between a conventional M module driver software architecture for the LXI equipment and hardware. In the method, an M module hardware interface control logic is realized by a field programmable gate array (FPGA); the LL driver layer driving of the M module software architecture is used for realizing an M module interface operating function; in the driving of a high-level (HL) driver layer, a function of an LL driver layer is called; a testing function related application layer repackages a function interface provided by the HL driver layer; a testing equipment related application layer realizes testing functions; the LL driver layer is divided into two architectures comprising a user model and a kernel model, the user model adopts a uClinux application program to perform FPGA setting and register configuration, and the kernel model adopts uClinum character driving to perform the setting of an advanced RISC machine (ARM) processor and the register configuration; and a system calls and interrupts corresponding functions of the user model and the kernel model.

The invention discloses a topic modeling based multi-granularity sentiment analysis method. The method includes the following steps: extracting words and word features of all data in a social media text database; performing training to obtain a kernel model; acquiring search results from the social media text database according to a search request of a user; determining number of topics and number of fine-grained sentiments needed in topic modeling according to parameters set by a system automatically or specified by the user; allocating one topic and one fine-grained sentiment to each word randomly; computing the topics and the fine-grained sentiments that all words belong to as well as coarse-grained sentiments expressed by searched documents, and feeding results back to the user. The method has the advantages that social network text data can be subjected to topic modeling and multi-granularity sentiment analysis at the same time; correlation can be established between the word features and the fine-grained sentiments expressed by the words, and the user is assisted in comprehending the data.

The invention provides an LSSVM (Least Square Support Vector Machine) pulsation wind speed prediction method based on a Morlet wavelet kernel. The prediction method comprises the following steps: utilizing an ARMA (Auto-Regressive and Moving Average) model to simulate and generate a vertical spatial point pulsation wind speed sample, dividing the pulsation wind speed sample of each spatial point into two parts including a training set and a test set, and carrying out normalization processing on the two parts; establishing an LSSVM model of the Morlet wavelet kernel; utilizing a Morlet wavelet kernel model subjected to PSO (Particle Swarm Optimization) to transform a pulsation wind speed training sample into a kernel function matrix, and mapping the kernel function matrix into a high-dimensional characteristic space; obtaining a nonlinear model of the pulsation wind speed training sample, and utilizing the model to predict the pulsation wind speed training sample; and comparing the wind sped of the test sample with a predicated pulsation wind speed, and calculating an average error, a root-mean-square error and a relevant coefficient of predicted wind speed and practical wind speed. The accuracy of pulsation wind speed prediction is guaranteed, and new wavelet kernel function selection with high precision and stability is provided.

The invention relates to an anti-kernel-hook method based on address randomization. The method comprises the following steps of copying an initialized kernel model to a randomly selected new memory position, and releasing an old memory space occupied by the kernel module; performing address relocation on the kernel module in the new position; correcting kernel module management information and registration information of the kernel module to a kernel sub system; and changing a normal registration function of the kernel module or an operation set address into a new memory position. The method has the advantages that rootkit can be discovered during the first execution of a hook function of the rootkit; an unknown rootkit kernel hook can be prevented; after the module initialization is completed, the random processing is added; partial kernel interface functions are modified; a registration function recording function is added; the performance expense is effectively reduced; the operating system kernel protection is guaranteed; and the security and the completeness of a system kernel are effectively enhanced.

The invention relates to a hyperspectral image classification multi-kernel learning method capable of maximizing the class separability, and relates to multi-kernel learning model solving. The invention aims to solve problems that multi-kernel model solving is not considered to be combined with subsequent classification application and the solving efficiency is low in an existing multi-kernel learning method applied to hyperspectral image classification. The hyperspectral image classification multi-kernel learning method is implemented according to the following steps: step one, training samples and test samples are acquired from a given input hyperspectral image data set; step two, a base kernel matrix Km in the multi-kernel learning model is constructed by using the training sample set Xtrain={xi}<i=1><N>, and a base kernel matrix set {Km}<m=1><M>={K1, K2,..., Km} is acquired; step three, the within-class discrete degree and the between-class discrete degree of the data set in a Hilbert kernel space are measured by using the base kernel matrix set; and step four, the class separability is measured according to a maximum class interval principle on the basis of the within-class discrete degree and the between-class discrete degree, and the weight of the base kernel is solved by taking the maximum class separability as a solving principle of the multi-kernel learning model. The hyperspectral image classification multi-kernel learning method is applied to the field of pattern recognition.

The present invention provides a multi-empirical kernel classifier based on Universum learning. Sample data is subjected to multi-empirical kernel mapping, and Universum sample data is generated in the sample space in each mapping. The present invention designs a Universum sample generation mode IMU (Imbalanced Modified Universum) which is independent of the model and can be utilized in other Universum learning. Imbalance of samples is introduced in the IMU to better solve the imbalance problem. A regularization item is designed according to the Universum sample, the regularization item is introduced into a multi-empirical kernel model to form a multi-empirical kernel classifier MUEKL (Multiple Universum-based Empirical Kernel Learning) based on the Universum learning. According to the generated target function, unknown parameters are subjected to optimization solution. Different from known multi-kernel learning, the multi-empirical kernel classifier combines the multi-empirical kernellearning and the Universum learning to provide an MUEKL algorithm; the expression on the balance data set is excellent through combination of the kernel learning and the Universum learning, and the imbalance problem is solved; and moreover, the present invention further provides a Universum data generation mode IMU which has a wide application.

The invention provides an iteration method and device for a kernel model, and the method comprises the steps: carrying out the desensitization of received data, and extracting and screening feature data from the desensitized data; selecting hyper-parameters from the feature data according to the state of the nuclear body model, and training the nuclear body model; and evaluating the trained nuclear body model, and after the evaluation is qualified, putting the iterated nuclear body model on line. The problem that the performance of the deep learning algorithm is influenced by data change because the distribution of the test data and the distribution of the training data are changed due to different time is solved; meanwhile, the problem that the performance of the deep learning algorithm is affected by data change due to the fact that the distribution of test data and the distribution of training data change due to different scenes is also solved; according to the method provided by the invention, the updated model can be adaptive to each scene, the extreme performance of the algorithm is brought into play, and the nuclear body verification accuracy of the algorithm is improved.

The invention discloses an ANA fluorescent sheet automatic identification method based on machine learning and deep learning, belongs to the technical field of machine learning and deep learning, andsolves the problems that the requirement on personnel is high and misjudgment is easy to occur when an ANA fluorescent sheet is judged manually in the prior art. The method includes based on a machinelearning model, obtaining a titer model through data set calculation; based on the deep learning model, obtaining a kernel model through data set calculation; reading a to-be-identified ANA fluorescent sheet, namely a to-be-identified picture, and performing feature extraction on the to-be-identified picture to obtain a plurality of extracted feature values; inputting the plurality of features into a titer model to obtain predicted titer; inputting the to-be-identified picture into the kernel model to obtain a predicted kernel; and obtaining an identification result according to the predictedtiter and the predicted karyotype. The method is used for identifying the titer and the karyotype of the ANA fluorescent sheet.

The present disclosure relates to an aligner in a hardware accelerator that can align short reads with a reference genome as genomic data is streamed through the hardware accelerator and thus can speed up the process of alignment. In an aspect, the disclosed short read aligner can incorporate a number of hardware kernels modelled as processor array implementation of the cost function model of the dynamic programming algorithm having a number of processing elements, wherein each kernel can incorporate a traceback control block as a separate hardware that enables traceback in parallel to the processor array and alignment matrix filling process by use of trace back direction vectors and using additional trackback path prediction features. The disclosed aligner can be parameterized and can perform alignment for cost function models of different variations of chosen dynamic programming algorithm. The aligner incorporates adequate sequence partitioning, scheduling, alignment and stitching schemes to accommodate short reads of variable lengths for alignment.

The invention relates to a dynamic load model modeling method adopting a support vector machine. The method comprises the following steps: collecting power grid failure data of a specific power branch of a certain power grid according to power grid data acquired in real time; summarizing data of a single-phase earth failure, an inter-phase short-circuit failure and a three-phase short-circuit failure, training load model data by the support vector machine to obtain a support vector machine model, and setting a structure of the model; acquiring an active power load difference equation model of the power branch according to a model equivalence principle; and acquiring a parameter theta and a parameter theta' of a system to be determined according to a linear kernel model and the structure of a system to be recognized. Through adoption of the dynamic load model modeling method, comprehensive consideration is given to experience risk minimization and structural risk minimization, so that an obtained power load model has higher generalization capability, can reflect the dynamic state of the system better, and is more approximate to a response feature of a real power system.

The invention discloses an efficient space k kernel mining method for space data. In order to find a space k kernel sub-graph meeting distance and cohesiveness constraints at the same time, the invention provides a new sub-graph model, namely a space k kernel model, on a space data set, and the new sub-graph model meets three conditions: the distance between two vertexes is not greater than a distance threshold r; each vertex has at least k neighbors; and any hyper-graph of the method cannot meet the former two conditions. Considering the attributes of the quadtree and the k core, the invention provides a new pruning strategy, so that the search space is reduced more effectively. Meanwhile, an efficient bound-based algorithm is developed in combination with a new pruning strategy, so thata spatial k core can be quickly found in a large spatial data set. Therefore, the application of the efficient spatial k-kernel mining method oriented to spatial data has great benefits for mining ofpotential communities and identification of key modes of spatial data sets.

The invention relates to a sandstone pore detection method based on multi-layer multi-kernel learning and region merging, and the method comprises the steps of obtaining an original pore image, carrying out the denoising and enhancement processing of the image, and carrying out the SLIC superpixel segmentation of the image; constructing an RAG region adjacency graph, and marking an adjacent relation between the regions; performing manual annotation after superpixel segmentation, and performing binary classification to obtain a pore region and a non-pore region; extracting artificial features from the training set images to obtain feature vectors of multi-feature fusion, and constructing a multi-layer multi-kernel model; extracting multi-feature vectors from the test set images, and inputting the multi-feature vectors into the stored model; outputting a probability that the current region is a pore, and taking the probability as a region similarity measurement value; calculating the similarity degree of each adjacent region in the graph; carrying out region merging; and outputting the pore detection area. The multi-layer multi-kernel learning algorithm is utilized to further improve the accuracy of region recognition, the target region and the background region of the image can be better distinguished, and the detection precision is improved.

The invention discloses a method for mining cohesion subgraphs in a bipartite graph network based on neighbor number and edge strength. In order to find a cohesion sub-graph simultaneously meeting theneighbor number and the edge strength, a new cohesion sub-graph model is provided on a bipartite graph network, namely a (alpha, beta, omega)-kernel model; two conditions are met: each node in the subgraph has enough neighbors with connected strong edges; wherein the strong edge represents that the edge is contained in at least omega butterflies, enough neighbors connected with the strong edge represent that at least alpha neighbors connected with the strong edge of the node exist in the upper layer of the bipartite graph, and at least beta neighbors connected with the strong edge of the nodeexist in the lower layer of the bipartite graph; wherein the sub-graphs are great. Considering the time complexity of calculating the model, the invention provides a new pruning strategy, so that thesearch space is reduced more effectively. Meanwhile, an efficient (alpha, beta, omega)-kernel model calculation algorithm is developed in combination with a new pruning strategy, so that the (alpha,beta, omega)-kernel model can be quickly found in the large bipartite graph network.

The invention discloses a lightweight traffic simulation system based on a springboot framework, and belongs to the technical field of calculation, reckoning or counting. The simulation system comprises an application display module, a simulation service module and a data management module. The application display module is used for system general situation description, road network data display, simulation task request and simulation result presentation; the simulation service module is used for simulation request analysis, simulation link response, simulation kernel model, basic data interaction, simulation result transmission and system log service; and the data management module is used for entity class establishment, mysql database maintenance, data interaction interface maintenance and database maintenance. According to the method, the simulation task request is sent through the visual front-end page, the execution link is matched for different simulation tasks, the corresponding kernel model and basic data are called, and finally the simulation result is returned to the request end and is analyzed and rendered, so that the development cost is effectively reduced, the development and deployment efficiency is improved, and the expansibility of the system is greatly enhanced.

The invention discloses a three-dimensional element image array coding method based on five-dimensional quadratic kernel modeling, and belongs to the technical field of computer image processing. The method comprises the following steps: deducing and establishing a theoretical system of a five-dimensional quadratic kernel model; executing a five-dimensional quadratic kernel hybrid expert algorithm initialized by five-dimensional Gaussian regression; executing a five-dimensional adaptive model selection algorithm; and executing the coding framework. The invention provides a coding method suitable for human eyes based on the theory of a five-dimensional quadratic kernel model and the characteristics of a three-dimensional element image. Model theory extension is carried out in application, and compared with a traditional HEVC and JPEG2000 coding algorithm based on transformation, the method has a good coding effect at a low bit rate compared with the two traditional coding methods.