48 results about "Matrix compression" patented technology

The invention relates to a composite material structure failure analysis method based on mesomechanics degradation model, and three dimensional basic material performances of the composite material are input parameters, and material performances of the composite material after damage are obtained by mesomechanics calculating method, and seven common failure modes are taken into consideration, which comprise: fiber stretch, fiber compression, matrix stretch, matrix compression, fiber-matrix shearing, layered stretch and layered compression, and finally the material degradation mode in composite material progressive damage analysis is obtained for carrying out failure analysis of the composite material structure; the invention not only can accurately predict structure failure, but also can predict failure mode and position of the structure. According to the composite material structure failure analysis method based on mesomechanics degradation model provided by the invention, the material degradation model is obtained by theory analysis, thereby connecting macroscopic failure and microscomic mechanism of composite material, and compared to the present material degradation model based on experience or test, time and test cost are substantially reduced.

This invention discloses one digital signal processor with bit expend and bit compression unit, which comprises digital signal processor nuclear, order memory and data memory, wherein, the nuclear comprises program control unit, order coding unit, address generation unit and data computation unit. This invention is characterized by the following: the computation unit is integrated with one bit expend and bit compression unit; the unit is composed of timer, displacement device and two source operation number and one aim register device.

A bioreactor device, and a method and system for fabricating tissues and growing cells and tissues in the bioreactor device, accommodates less than about 1 mL (or less than about 200 μL) of local medium volume but sample sizes of about 100 μL or greater. The bioreactor device includes a bioreactor chamber for containing a sample, where sample growth in response to mechanical, electrical, and biofactor stimulation is monitored through one or more optical ports. Embedded sensors are provided for measuring fluid pressure, pH, temperature, and oxygen tension. The bioreactor device can receive different types of mechanical loadings, including fluid shear, hydrostatic pressure, matrix compression, and clinorotation.

Associative matrix compression methods, systems, computer program products and data structures compress an association matrix that contains counts that indicate associations among pairs of attributes. Selective bit plane representations of those selected segments of the association matrix that have at least one count is performed, to allow compression. More specifically, a set of segments is generated, a respective one of which defines a subset, greater than one, of the pairs of attributes. Selective identifications of those segments that have at least one count are stored. The at least one count that is associated with a respective identified segment is also stored as at least one bit plane representation. The at least one bit plane representation identifies a value of the at least one associated count for a bit position of the count that corresponds to the associated bit plane.

This invention relates to utilising image compression and reconstruction algorithms for transmission and reception of data that can be represented as a numerical matrix or a sequence of numerical matrices in bandwidth limited channels. The method and algorithms described herein may reduce data size to such an extent that data that was considered too large to be practically transmitted in a useful time frame can be transmitted in real or near real time. The algorithms can be applied to data communications for a variety of sensor, electronic and information systems.

The invention provides a depth learning model matrix compression method and device. The last linear layer of a depth learning model is connected with M hidden nodes and N classification nodes. The last linear layer is corresponding to a weight matrix W. The method comprises the steps that S101 according to the absolute values of the elements of the weight matrix W, a K value is calculated; and Step S102 the last linear layer is divided into a first linear layer and a second linear layer, wherein the weight matrix of the first linear layer is a matrix P of M*K, and the weight matrix of the second linear layer is a matrix Q of K*N. The output of the first linear layer is the input of the second linear layer. M*N is greater than K*(M+N), and the weight matrix W is compressed.

The invention discloses a method for calculating a complex structure radiation sound field in a marine acoustic channel, and belongs to the technical field of acoustic numerical calculation. The limitation that a traditional boundary element method is only suitable for free field sound field calculation is broken through, and a sound field Green function in an ocean sound channel is established. Ahierarchical matrix compression technology is adopted to divide a radiation impedance matrix into a series of hierarchical matrix blocks with different sizes, and an iteration method is further utilized to solve a matrix equation. Compared with the prior art, the method has the effects and benefits that the acoustic calculation function of the traditional boundary element method is expanded frominfinite uniform media to the bounded space of the ocean channel, and the influence of sea surface, seabed boundary and sound velocity gradient distribution can be considered. And on the other hand, the radiation impedance matrix is subjected to hierarchical compression, the consumption of a computer memory is greatly reduced by utilizing the low-rank characteristic approximate decomposition of the matrix block, the influence of the diversity of a sound field Green function under the complex marine channel condition is avoided, and the algorithm adaptability is good.

The invention discloses a code classification and clustering method based on convolution and a recurrent neural network. An experimental data set comprises 104 types of codes. The specific steps of code classification are as follows: generating AST of codes by using srcml,generating a word vector of the node by using Word2vec according to the expanded sequence of the AST, cutting the AST into a subtree sequence composed of a cycle and a condition AST, wherein each sub-tree uses a tree convolutional neural network TBCNN for encoding, inputting a vector obtained by encoding into a bidirectionalLSTM, putting a code vector generated in each time step into a matrix, compressing the matrix by using maximum pooling to obtain a vector, mapping the vector by using a layer of fully connected neuralnetwork to obtain a 104-dimensional code vector, generating a marker vector by using a onehot method, wherein cross entropy is used as a loss function, carrying out dimension reduction by using a TSNE method, and clustering on the codes by using a Kmeans clustering method. The classification accuracy can reach 97.8%, and the clustering accuracy can reach 83.6%.

Associative matrix compression methods, systems, computer program products and data structures compress an association matrix that contains counts that indicate associations among pairs of attributes. Selective bit plane representations of those selected segments of the association matrix that have at least one count is performed, to allow compression. More specifically, a set of segments is generated, a respective one of which defines a subset, greater than one, of the pairs of attributes. Selective identifications of those segments that have at least one count are stored. The at least one count that is associated with a respective identified segment is also stored as at least one bit plane representation. The at least one bit plane representation identifies a value of the at least one associated count for a bit position of the count that corresponds to the associated bit plane.

The invention discloses a coal pore correction method based on a mercury injection experiment. The coal pore correction method comprises the following steps: 1) performing low pressure section and high pressure section mercury injection tests on a to-be-tested coal sample by utilizing a mercury injection instrument and obtaining mercury injection data; 2) performing pore structure correction including matrix compression effect correction, inter-granular pore effect and skin effect correction effect correction and pore shape effect correction; 3) performing quantitative characterization on a pore structure by utilizing corrected data. The invention provides a model for correcting coal pores by utilizing the mercury injection tests, which can be used for coal-bed methane reservoir evaluationand research on gas prevention and control and the like; a mercury injection test method in the national standard is referenced and utilized, mercury injection pressure and mercury injection amount data are obtained, and then related factors affecting the mercury injection test accuracy are corrected; correction results can accurately characterize the coal pores. Therefore, a strong technical support is provided for coal-bed methane exploitation and underground gas prevention and control.

The invention belongs to the technical field of data mining, and specifically provides a text semantic representation method based on an aggregation weighting matrix compression algorithm. According to the method, through constructing a global smooth context matrix, influences of pragmatic habits on word semantics are weakened; through utilizing a word vector aggregation algorithm to weight global smooth contexts, a word vector weight measurement method is corrected; and finally, texts are represented through utilizing an infinite norm compression algorithm to convert global smooth contexts into vectors. Compared with the existing method, the method has the advantages that the highest possible value of each dimension of the vectors is reserved and the text information is represented more fully.

The invention relates to a method for acquiring the electric field distribution of a large-sized complex aircraft by using a fast boundary element method, which is technically characterized by solving the problem of the fast computation and analysis of a large-sized electrostatic field of the aircraft in the field of aircraft design by adopting the boundary element method. The method comprises the following steps of: dividing a medium interface into smooth curved surfaces with parametric equations by adopting a Nessler discretization method, and projecting integral points to the medium interface to form boundary integral points through the parametric equations; and constructing a wavelet transform matrix on the basis of the multiscale grouping of the integral points, and carrying out matrix compression, computation and storage on a boundary element coefficient matrix. The invention overcomes the shortages of complex computation, large computation amount and low precision of the coefficient matrix in the traditional wavelet boundary element method and effectively reduces the internal storage occupying amount and the computation time of boundary element analysis, thereby solving theproblems of a large-scale project with millions of freedom degrees; and in addition, the invention can be applied to the electric field computation and analysis of the large-sized complex aircraft inan aerospace field and electric field analysis in other project fields.

The invention discloses a base vector matrix compression device based on a high-speed systolic array and Givens transformation. The base vector matrix compression device is applied to a base vector matrix V with the dimensionality of M*N (M>=N) and comprises a preprocessing module and a compression module, wherein the preprocessing module is composed of a first-kind preprocessing unit and second-kind preprocessing units; the compression module is composed of real number processing units, Givens rotation processing units and simplified Givens rotation processing units. The base vector matrix compression device can guarantee the regularity of hardware implementation and parallel flow processing of data, is high in universality and can achieve compression of any base vector matrix with the dimensionality of M*N.

Apparatus and method for context-aware compression. For example, one embodiment of an apparatus comprises: ray traversal / intersection circuitry to traverse rays through a hierarchical acceleration data structure to identify intersections between rays and primitives of a graphics scene; matrix compression circuitry / logic to compress hierarchical transformation matrices to generate compressed hierarchical transformation matrices by quantizing N-bit floating point data elements associated with child transforms of the hierarchical transformation matrices to variable-bit floating point numbers or integers comprising offsets from a parent transform of the child transform; and an instance processor to generate a plurality of instances of one or more base geometric objects in accordance with the compressed hierarchical transformation matrices.

The invention provides a precondition conjugate gradient block adjustment method based on a sparse block matrix compression storage structure. A large-scale sparse matrix compression storage algorithm is adopted, a large-scale method equation generated in big data block adjustment is subjected to compression storage and operation, the demand for memory is effectively reduced, a precondition conjugate gradient method iterative solution method equation is introduced aiming at an inversion problem of a compression matrix, direct inversion of the compression matrix is avoided, the storage and operation problem for the large-scale method equation in traditional block adjustment is solved, and the data processing capacity and efficiency are improved. The method is especially suitable for block adjustment of middle-large testing zone (500-10,000 images) data.

The invention provides an unsigned approximate multiplier with low logic complexity. The approximate multiplier comprises a partial product matrix generation module, a first-stage approximate compression module, a second-stage approximate compression module, a second-stage accurate compression module and an adder module. Approximation is introduced into the compressed part of the partial product matrix, and an approximate compressor is completely used in the first stage of compression of the partial product matrix; in the second stage of partial product matrix compression, a full adder and a half adder are used in the high-order weight part, and only an approximate compressor is used in the low-order weight part. The approximate compressors used in the method only use logic AND gates and logic OR gates, and compared with accurate compressors (a half adder and a full adder), the logic is simple, and hardware expenditure is small. According to the approximate multiplier, the accuracy isguaranteed while the circuit performance is improved, and good balance is achieved in the aspects of precision and hardware cost.

The invention provides a straw and mushroom matrix compression forming machine and belongs to the technical field of cultivation of edible mushrooms. The straw and mushroom matrix compression forming machine mainly comprises a bed, a discharging device, a compactor and a die. A raw material hopper and the discharging device part are arranged at the head part of one end of the bed. The compactor part is arranged at the tail part of the other end of the bed. The die capable of reciprocating is arranged at the middle part of the bed. The straw and mushroom matrix compression forming machine has simple and compact structure and is easy to operate. Straw culture materials can be compressed into one fifth of the original volume of the straw culture materials, so that the output in unit area is possibly improved for over five times; and compressed material blocks can be stacked and is developed toward the space and a mushroom bed does not need to be set up.

The invention discloses a method for testing a triple pore deformation coupling rule of coal rock in an unbalanced state, and relates to the technical field of coal rock deformation testing. The method comprises the following steps: firstly, crushing and processing raw coal into a briquette coal sample with a crack in the middle; then, pasting a strain gauge on the surface of the briquette coal sample, covering all the surfaces of the briquette coal sample with a soft silicone rubber adhesive, and reserving gas injection ports at the two ends of the crack; then, putting the briquette coal sample into a pressure container, and integrally putting the pressure container into a constant-temperature chamber; and finally, after the temperature of the constant-temperature chamber is stable, injecting gas, and meanwhile, collecting data of the strain gauge, and measuring strain evolution caused by matrix compression or expansion of the briquette coal sample under the action of gas adsorption. According to the method, the deformation rule of different parts of the matrix of the briquette coal sample under the interaction among the multiple pore systems can be directly observed, and the permeability evolution rule of the multiple pore systems can be inverted by combining numerical simulation.

The invention discloses an electromagnetic characteristic analysis method and system for an electrically large-size target with a complex structure, and the method comprises the steps: carrying out the surface triangulation of the target, setting a regional decomposition seed, completing the regional autonomous decomposition of the target based on the decomposition seed, obtaining a plurality of decomposition parts, and numbering the decomposition parts; creating a plurality of parallel threads, specifying a main thread and sub-threads, and distributing the number and serial number of decomposition parts responsible for each sub-thread; for each sub-thread, constructing a comprehensive function for all responsible decomposition parts, compressing the self / mutual impedance matrix and the excitation matrix of the decomposition parts by utilizing the comprehensive function, and reporting a result to a main thread; the main thread detects connection boundaries between different decomposition components, and calculates a self / mutual impedance matrix and an excitation matrix of the connection boundaries; and constructing a target overall compression matrix equation, solving the equation,and combining with the comprehensive function definition to obtain the electromagnetic characteristics of the target in space. The method solves the problems of low processing efficiency and the likein the prior art, and greatly improves the comprehensive processing efficiency of the SBFM.

The invention discloses a neural network compression method using a block cyclic matrix. The invention relates to the field of neural network compression. The method comprises the following steps: reading a longest basic random symbol vector in a neural network, generating a symbol vector equal to an input dimension of the layer on each layer of the neural network, multiplying the symbol vector byan input vector element to obtain a new input vector, training to form a new block cycle network, and storing the longest basic random symbol vector and neural network model parameters; and pruning the model by adopting a fine-grained neural network compression method, and further reducing the complexity of the model. By introducing the random symbol vectors, the correlation between the projection vectors is reduced, so that the convergence of the model is ensured, and the purpose of effectively reducing storage and bandwidth is achieved. And meanwhile, the problem of performance reduction caused by increase of the block size when the block cyclic matrix compression neural network is processed is avoided.

Associative matrix compression methods, systems, computer program products and data structures compress an association matrix that contains counts that indicate associations among pairs of attributes. Selective bit plane representations of those selected segments of the association matrix that have at least one count is performed, to allow compression. More specifically, a set of segments is generated, a respective one of which defines a subset, greater than one, of the pairs of attributes. Selective identifications of those segments that have at least one count are stored. The at least one count that is associated with a respective identified segment is also stored as at least one bit plane representation. The at least one bit plane representation identifies a value of the at least one associated count for a bit position of the count that corresponds to the associated bit plane.

The invention discloses a word vector matrix compression method and device and a word vector acquisition method and device, which relate to the technical field of data processing, and are beneficial to saving the storage space of client equipment. The word vector matrix compression method comprises the following steps of generating a word vector matrix for representing a to-be-processed vocabularybased on a word vector model, and taking the generated word vector matrix as a to-be-compressed word vector matrix, wherein one row or one column of the to-be-compressed word vector matrix is a wordvector, and one word vector in the to-be-compressed word vector matrix is used for representing one word in the to-be-processed vocabulary; classifying the word vectors included in the to-be-compressed word vector matrix according to the semantic information of the to-be-processed vocabulary to obtain at least two categories; and compressing at least one of the at least two categories, constructing a compressed word vector matrix according to the word vector obtained by compressing the at least one category, and storing the compressed word vector matrix.

The invention relates to a method for rapidly calculating parameters of a multi-port equivalent network element with embedded electromechanical transient simulation, and belongs to the technical field of electromechanical-electromagnetic transient hybrid simulation analysis of a power system. The method comprises the steps of determining the topology of a power network, based on a normal operation state or fault state of the power grid, forming a network node voltage equation matrix--an admittance matrix Y, and factorizing the matrix Y to obtain a result matrix A storing factorization; solving upper and lower triangular parts of the matrix A respectively to obtain a solution vector, then reconstructing a compression impedance matrix M and obtaining an inverse matrix compression admittance matrix E; according to elements of the matrix E, forming an equivalent multi-port network containing parameters and corresponding to a hybrid simulation sub-network interface actual topology; and obtaining the self impedance and mutual impedance of the actual topology in the electromechanical-electromagnetic hybrid simulation interface equivalence, and ending the calculation. The calculation process is simple, the result is accurate, and the efficiency and calculation speed of the calculation method are greatly improved.

The invention provides a point cloud compression method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining a to-be-compressed point cloud; the to-be-compressed point cloud is divided into a plurality of sub-point cloud blocks, the sub-point cloud blocks are described through a graph, and a weight matrix corresponding to the sub-point cloud blocks is determined; dividing the graph into a plurality of sub-graphs by adopting a spectral clustering algorithm; iteratively combining the sub-graphs according to a preset cost function, and determining optimized graphs corresponding to the sub-graphs; determining a target adjacent matrix corresponding to the optimization graph, and updating the weight matrix according to the target adjacent matrix; and determining a Laplacian matrix corresponding to the optimization graph according to the updated weight matrix, and compressing the to-be-compressed point cloud according to the Laplacian matrix. The compression performance and the compression effect of point cloud attribute compression can be improved.

The invention discloses a connected piece identification method based on association matrix compression and branch pointer vector update, which solves the problem of how to efficiently and reliably perform network connected piece identification. The invention comprises the following steps: step 1, generating a branch-node association matrix of the network according to the network branch-node related information; step 2, searching the zero-column of the branch-node association matrix to identify an isolated node; step 3, "compressing" the branch-node association matrix by rows, leaving only thecolumn labels of non-zero elements to obtain a plurality of binary connected sets; step 4, performing column scan on the branch-node association matrix and updating the branch pointer vector; and step 5, merging and growing the binary connected set according to the final branch pointer vector to obtain the final connected slice set. The method avoids a large number of graph search and logic operations, with no need to perform matrix decomposition operation, and is efficient and reliable to be suitable for advantages of accelerating by using sparse technology, etc.

Embodiments of the present invention provide a codebook information feedback method, a terminal and network equipment, the method comprising: feeding back codebook information of a CSI codebook, wherein the CSI codebook comprises a spatial beam vector matrix, a compression coefficient matrix and at least one compression vector, an element in the compression coefficient matrix is a quantized valueof a compression coefficient, and the compression coefficient is a coefficient obtained by compressing a combined coefficient or the compression coefficient is a coefficient obtained by compressing asampling combination coefficient, and the sampling combination coefficient is a combination coefficient of partial frequency domain granularity or the compression coefficient comprises a coefficient obtained by compressing the combined coefficient and then sampling, and the codebook information comprises spatial beam vector information contained in the spatial beam vector matrix, elements of the compression coefficient matrix and vector information of the at least one compression vector. According to the embodiment of the invention, the feedback overhead of the CSI codebook can be reduced.

The invention discloses a multi-layer matrix compression method for analyzing the mutual coupling of microstrip packaging interconnection lines. The steps are: extract the grid file of microstrip packaged interconnection lines, mark the interconnection line ports of mutual coupling; perform octree grouping on the analyzed interconnection line grid according to the number of discrete edges in the finest layer group, and group The distance between the centers divides the near-field action area and the far-field action area of ​​each group containing discrete edges; establishes the mixed potential integral equation and the right excitation vector corresponding to the interconnection line feed, and for the dense matrix generated by the integral equation discretization, The moment method is used in the near-field action area, and the multi-layer matrix compression method is used in the far-field action area; the current coefficient of the equation is determined by the GMRES iteration method of ILU preconditioning, and the relationship between the interconnection lines is solved according to the marked interconnection lines. coupling. The method of the invention has the advantages of less memory consumption and high efficiency, and realizes fast and effective estimation of the mutual coupling of the entire microstrip packaging interconnection line.

The invention belongs to the technical field of data mining, and specifically provides a text semantic representation method based on an aggregation weighting matrix compression algorithm. According to the method, through constructing a global smooth context matrix, influences of pragmatic habits on word semantics are weakened; through utilizing a word vector aggregation algorithm to weight global smooth contexts, a word vector weight measurement method is corrected; and finally, texts are represented through utilizing an infinite norm compression algorithm to convert global smooth contexts into vectors. Compared with the existing method, the method has the advantages that the highest possible value of each dimension of the vectors is reserved and the text information is represented more fully.

The invention discloses a matrix compression algorithm for calculating the shortest path of an expressway. The matrix compression algorithm comprises the following steps of: 1, calculating a matrix graph of the path with the minimum cost of the expressway by using a Dijkstra algorithm; 2, performing matrix row compression, establishing a row tag array Flag [N] (initialized as full false), traversing a preorder matrix row by row, and compressing each row in the Pre array into a two-tuple set RowListi (column number c, value v); 3, matrix column compression: traversing the matrix column by column, compressing each column in the Pre array into a triple set ColListk (a starting value s, an end value e and a value v), and compressing and neglecting special points and repeated points in the matrix; and 4, completing matrix compression, and compressing the preorder array through a matrix compression algorithm, so that the required memory is greatly reduced, and the retrieval efficiency is not reduced too much.