249 results about "Matrix element" patented technology

The invention discloses an ore element grade detection method. The method comprises the following steps: respectively obtaining intensity values of X-fluorescence rays absorbed by interference elements and elements to be detected in an ore sample to be detected; inputting the intensity value of the X-ray absorbed by the to-be-detected element into a pre-trained preliminary grade value prediction model to obtain a preliminary grade value of the to-be-detected element; the preliminary grade value prediction model is constructed by using a unary linear regression method; inputting the intensity values of the X fluorescence rays absorbed by the interference elements and the to-be-detected elements into a pre-trained correction compensation model to obtain correction compensation values of the to-be-detected elements; the correction compensation model is constructed by using a dynamic regularization random weight neural network; and taking a sum value of the preliminary grade value and the correction compensation value as a final grade estimation value of the to-be-detected element, so as to obtain a detection result of the to-be-detected ore element grade. According to the method, errors caused by mutual interference among matrix elements are effectively analyzed in an ore dressing field environment, and the accuracy of a detection result is improved.

A hardware accelerator including tiles arranged in a systolic array. At each of the tiles, the systolic array receives a first input block that includes first input matrix elements of a first input matrix. In each of a plurality of multiplication iterations, at each of the tiles, the systolic array receives a respective second input block. The systolic array computes tile products of the first input matrix elements and second input matrix elements included in the second input blocks. The systolic array adds the tile products to column-wise partial sums and transmits the column-wise partial sums to subsequent tiles along accumulator rings included in array columns of the systolic array. In a subset of the multiplication iterations, the systolic array outputs product block rows of a product matrix. The product block rows each include product matrix blocks computed as rows of the column-wise partial sums.

The invention relates to the technical field of electrical digital data processing, in particular to a hardware test resource allocation method based on dynamic topology adaptation, which comprises the following steps of: 1, constructing a dynamic topology sensing network: acquiring electrical connection signals between devices on a test backboard in real time, generating and periodically updating a topology connection matrix, and establishing a dynamic topology sensing network; the matrix element value represents the on-off state of a physical link between devices, and when a device plugging event is detected, matrix updating is triggered immediately; 2, quantifying resource and task attributes; 3, dynamic resource matching of topology driving is executed; and 4, responding to non-interruption migration of topology change: when the topology connection matrix is updated, executing the step 3 again on the influenced task, and realizing non-interruption switching through new and old resource parallel initialization and signal deviation fault-tolerant verification. Through the dynamic topology adaptation technology, test resources can be dynamically adjusted and matched according to real-time topology and task requirements, and efficient utilization of the resources is ensured.

Systems or methods of the present disclosure may provide an integrated circuit system that includes a programmable logic device that includes a clock, one or more local controllers, programmable logic units implementing a systolic array to compute a matrix multiplication, and embedded memory blocks. The embedded memory blocks include a single port random access memory (SPRAM). The one or more local controllers are configured to, on a first set of alternating clock cycles of the clock, load matrix sub-elements from two rows of a matrix into corresponding matrix element of the SPRAM. The one or more local controllers are configured to, on a second set of alternating clock cycles of the clock, read out the matrix elements from the SPRAM to the systolic array to compute the matrix multiplication.

The invention provides a data transmission method and apparatus, an electronic device and a computer readable storage medium. The method comprises the steps of obtaining a data transmission request sent to a port of a router node; based on a matrix row in a polling matrix matched with the router node, performing polling arbitration on the data transmission request, and determining that a target port of the router node processes data of the data transmission request; updating matrix elements in a matrix row and a matrix column corresponding to the target port, and performing next arbitration; wherein each matrix row represents a port of the router node in one transmission direction; different matrix rows represent different transmission directions, and the matrix elements in each matrix row are used for reflecting the blocking condition of the port corresponding to the matrix row. According to the invention, the arbitration efficiency can be improved, and the stability and fairness of the on-chip internet are improved.

The invention provides a video processing method and device and electronic equipment, and the method comprises the steps: inputting a target video into a coding model, and determining the potential spatial features of each video frame through the coding model; constructing a similarity kernel matrix based on the potential spatial features; determining candidate sub-matrixes from the similarity kernel matrix, and determining the candidate sub-matrix with the maximum element difference among matrix elements in the candidate sub-matrixes as a target sub-matrix; and carrying out dimension reduction processing on the target sub-matrix, and outputting compressed data of the target video based on a dimension reduction result through a decoding model. In the mode, based on the potential spatial features of the target video determined by the encoder, the similar kernel matrix is constructed, representative samples are selected from the similar kernel matrix, dimension reduction is carried out on the samples, and the dimension reduction result is reconstructed again through the decoder, so that more redundant information can be removed on the premise of retaining data information, and the accuracy of the data information is improved. And data processing is performed in the potential space, so that the calculation cost is reduced, and the requirements of reducing storage and calculation resources are met.

Disclosed is an operation device. The operation device includes a memory storing at least one instruction; and a processor configured to execute the at least one instruction, and the processor, by executing the at least one instruction, may perform encoding or decoding for an approximate homomorphic ciphertext using a predetermined matrix having only a half of an element of a matrix corresponding to a canonical embedding function.

The present invention discloses a reconfigurable integrated optical computing core model, comprising: M×N matrix calculation units; each matrix calculation unit includes two dual-waveguide microring resonators, a straight waveguide, and a phase change material heater, and the straight waveguide is made by covering a phase change material on a silicon waveguide; the phase change material heater is used to heat the phase change material to generate a phase change, thereby changing the light absorption of the straight waveguide; the integrated optical computing core model realizes the multiplication operation of an M×N matrix and an N×1 vector through integrated optical methods, where M and N are natural numbers. The present invention uses a phase change material in the computing core, changes the phase state of the phase change material by heating, changes the light absorption of the straight waveguide, and further changes the matrix elements in the computing core, realizing the reconfigurable characteristics of the computing core model. The present invention can be widely applied to the field of integrated optics.

The invention provides a production line parameter real-time scheduling method based on reinforcement learning, and belongs to the technical field of production lines, and the method comprises the steps: collecting sensor data to form an original state vector, reducing the dimension of the original state vector into a low-dimensional feature state vector through a state compression encoder, inputting the low-dimensional feature state vector corresponding to a microcosmic scheduling unit into a parameter decision model, and carrying out the real-time scheduling of the microcosmic scheduling unit; the attention weight coefficient of the model is determined by the product of the historical scheduling success rate, the element value of the difference degree matrix and the maximum characteristic value of the inter-stage sensitivity matrix, after a scheduling instruction is output, simulation evaluation is performed in a digital twin platform, and after the scheduling instruction passes, a real production line is issued for execution and multi-time-scale deviation indexes are collected; and calculating a comprehensive reward value according to the indexes, and storing an experience sample to an experience playback buffer area for model online update training, thereby solving the technical problem that production line parameter scheduling is difficult to consider multi-level state feature recognition and dynamic decision weight optimization at the same time.

A processing apparatus is described herein that includes a general-purpose parallel processing engine comprising a matrix accelerator including one or more systolic arrays, at least one of the one or more systolic arrays comprising multiple pipeline stages, each pipeline stage of the multiple pipeline stages including multiple processing elements, the multiple processing elements configured to perform processing operations on input matrix elements based on output sparsity metadata. The output sparsity metadata indicates to the multiple processing elements to bypass multiplication for a first row of elements of a second matrix and multiply a second row of elements of the second matrix with a column of matrix elements of a first matrix.

Embodiments of apparatuses and methods for copying and operating on matrix elements are described. In embodiments, an apparatus includes a hardware instruction decoder to decode a single instruction and execution circuitry, coupled to hardware instruction decoder, to perform one or more operations corresponding to the single instruction. The single instruction has a first operand to reference a base address of a first representation of a source matrix and a second operand to reference a base address of second representation of a destination matrix. The one or more operations include copying elements of the source matrix to corresponding locations in the destination matrix and filling empty elements of the destination matrix with a single value.

A computing system for performing runtime data handling optimization for generative models is provided. The computing system comprises at least one processor and memory comprising a first memory and a second memory, wherein the memory stores instructions that, when executed by the at least one processor, cause the at least one processor to execute a generative model. The computing system computes a first value matrix entry based upon the processing of an input to the generative model. The first value matrix entry is stored in a first memory wherein a first group of value matrix entries is identified. The computing system executes data quantization on the first group of value matrix entries which results in a first quantized value matrix. The first quantized value matrix is added to a second memory where it can be used during generation of the generative model. When the first group of value matrix entries is less than a group size parameter, data padding matrix values are generated and used during execution of the data quantization.

A sparse dot product and / or matrix multiply is computed by subdividing each vector and simultaneously performing operations to generate output matrix elements. In an embodiment, a bit mask is computed that includes one bit for each element of an input matrix, each bit indicating whether the element is non-zero or zero. In an embodiment, the element values are stored in a packed format, where all of the non-zero values are packed together and the remaining storage for the matrix contains zeros (or any other values). The bit mask can then be used to determine the location of each non-zero element in the packed storage. Rather than reading all of the elements, only the non-zero elements that will be multiplied by a non-zero element from the other input vector should be read. Any multiplication by a zero element from either input vector or matrix is unnecessary.

The invention relates to 750MPa-grade high-strength steel with high comprehensive formability and a preparation method, and belongs to the technical field of steel preparation. The high-strength steel comprises the following chemical components in percentage by mass: 0.04%-0.08% of C, less than or equal to 0.2% of Si, 1.4%-2.0% of Mn, less than or equal to 0.01% of P, less than or equal to 0.005% of S, 0.02%-0.05% of Al, 0.02%-0.06% of Nb, 0.08%-0.15% of Ti, 0.1%-0.3% of Mo, less than or equal to 0.004% of N and a matrix element Fe, the microstructure of the high-strength steel comprises, by volume fraction, 60%-80% of ferrite and 20%-40% of bainite. Through reasonable component design, a metallographic structure of a target proportion is obtained, so that the obtained high-strength steel has relatively high elongation and hole expansion rate, and has excellent comprehensive forming performance.

A Winograd algorithm multiplies a first matrix by a second matrix to generate a third matrix. A plurality of second accumulated value calculation units, accumulates second multiplication values obtained by multiplying each of paired element values of the second matrix, a second accumulated value output unit outputting selecting and outputting one of output values of adjacent second accumulated value calculation unit and an accumulated second multiplication value as a second accumulated value, a third accumulated value output unit including a plurality of third accumulated value calculation units and generating third accumulated value, and one or more row element value calculation units, of which one row element value calculation unit accumulates first matrix element multiplication values obtained by multiplying each of the paired element values of the first matrix, and calculates element values of the third matrix using the accumulated first matrix element multiplication value and the second accumulated value.

The disclosed embodiments relate to matrix compression / decompression instructions. In one example, a processor includes fetch circuitry to fetch a compression instruction having a format with fields to specify positions of an opcode and a decompressed source matrix and a compressed destination matrix; the decoding circuit is used for decoding the extracted compression instruction; and execution circuitry to, in response to the decoded compression instruction, generate a compressed result according to a compression algorithm by compressing a specified decompressed source matrix via any of: compress non-zero-value elements together and store a matrix position of each non-zero-value element in a header, or using fewer bits to represent one or more elements and using a header to identify matrix elements represented by fewer bits; and storing the compressed result to the specified compressed destination matrix.

The application relates to a power data fusion method, comprising the following steps: acquiring time series data from multiple data sources respectively; calculating the similarity between each time series data to obtain a similarity matrix; accumulating the matrix elements in the same row or the same column in the similarity matrix to obtain the similarity accumulation value of each data source and other data sources; performing longitudinal parameter fusion on each data source with a similarity accumulation value higher than a first threshold value, comprising the following steps: performing data preprocessing on the time series data; calculating the grey correlation degree between each two time series data of the time series data after the data preprocessing; filtering out the time series data with a grey correlation degree lower than a third threshold value; performing data fusion on the time series data by using a BP neural network to obtain a first fusion result; and performing horizontal parameter fusion on each data source with a similarity accumulation value lower than the first threshold value.

The present invention proposes a method for solving the transmission characteristics of a cellular frequency selective surface. The implementation steps include: constructing a unit geometric model of the cellular frequency selective surface; performing mesh division on the unit geometric model; calculating the direction vector of the polarized plane wave in OXYZ; calculating the incident field at the excitation port in different modes; constructing a finite element matrix equation; performing matrix element substitution on the matrix equation; and obtaining the transmission characteristics of the cellular frequency selective surface. By using two vector edges of the excitation port and transforming the three-dimensional coordinate system into a three-dimensional local coordinate system, and calculating the wave numbers of the polarized plane wave in the three axial directions of the three-dimensional local coordinate system, the present invention calculates the incident field at the excitation port in different modes, avoiding the defect in the prior art that the incident field at the excitation port cannot be calculated in a coordinate system with two vector edges of a regular hexagon as the coordinate axes by using the finite element method, and achieving high-precision solution of the transmission characteristics of the cellular frequency selective surface.

This method, belonging to the field of electromagnetic engineering, aims to improve the speed of impedance matrix filling when solving electromagnetic scattering fields of rotating bodies. Based on a higher-order Taylor series method, it obtains more accurate impedance matrix elements with less sampling data, thus significantly saving computational resources and improving design efficiency. Simulations of various rotating electromagnetic devices demonstrate the superior performance in terms of impedance matrix filling accuracy, speed, and memory consumption, as well as its universality for different models. The method derives the modal Green's function expression after higher-order integration and performs simulations using specific electromagnetic rotating device models. The increased order of the method has practical application value in improving the accuracy of device model simulations. The higher-order numerical integration method also brings efficiency improvements, saving memory consumption and significantly increasing the impedance matrix filling speed compared to the traditional T-shaped integration method.

The invention relates to a low-carbon-emission high-performance IF steel plate and a preparation method, and belongs to the technical field of steel preparation. The IF steel plate comprises the following chemical components in percentage by mass: less than or equal to 0.0030% of C, less than or equal to 0.0060% of S, less than or equal to 0.030% of Si, less than or equal to 0.2% of Mn, 0.0030%-0.006% of N, 0.060%-0.12% of Al, 0.030%-0.080% of Ti, 0.0015%-0.030% of Nb, less than or equal to 0.0008% of B, less than or equal to 0.12% of Cu, less than or equal to 0.01% of Sn, less than or equal to 0.15% of Cr, less than or equal to 0.08% of Ni, 0.001%-0.02% of La + Se and a matrix element Fe, al / N is greater than or equal to 18 and less than or equal to 22, Ti is greater than or equal to 3.4 N, Nb is greater than or equal to 7.75 C, and Al / N represents the mass fraction ratio of Al to N. By accurately controlling the content of elements in steel, formation of a precipitated phase and distribution of grain boundary elements are controlled, and the influence of residual elements in waste steel and electric furnace nitrogen increase on the performance of a finished product is reduced. The short-process production of IF steel by the full-scrap steel electric furnace is realized, the production cost and carbon emission are reduced, and meanwhile, the high performance of the material is ensured.

The present application relates to the technical field of electric digital data processing, and especially relates to a hardware test resource allocation method based on dynamic topology adaptation, which comprises the following steps: step 1: constructing a dynamic topology-aware network; collecting electrical connection signals between devices on a test backplane in real time, generating and periodically updating a topology connection matrix, wherein the matrix element value represents the on-off state of the physical link between devices, and the matrix is updated immediately when a device plugging event is detected; step 2: quantifying resource and task attributes; step 3: performing topology-driven dynamic resource matching; step 4: responding to topology changes for non-interruption migration; when the topology connection matrix is updated, step 3 is re-executed for the affected tasks, and non-interruption switching is realized through parallel initialization of new and old resources and signal deviation fault tolerance verification. Through the dynamic topology adaptation technology of the present application, test resources can be dynamically adjusted and matched according to real-time topology and task requirements, ensuring efficient use of resources.

The invention discloses a vibration signal processing method, device and equipment and a computer readable storage medium, and belongs to the technical field of signal processing. The method comprises the steps that an original vibration signal matrix is obtained, the original vibration signal matrix comprises a plurality of first matrix elements, and any first matrix element comprises a vibration signal obtained by any sampling of any sampling point at any time; the original vibration signal matrix is processed, a representative vibration signal matrix is obtained, the representative vibration signal matrix comprises a plurality of second matrix elements, and any second matrix element comprises a vibration signal of any sampling point at any time; determining a third matrix element in the second matrix elements; determining a fourth matrix element in the original vibration signal matrix according to the sampling point and time corresponding to the third matrix element; and according to the sampling point, the time and the vibration signal corresponding to the fourth matrix element, determining a fifth matrix element meeting the condition in the fourth matrix element. The method improves the processing efficiency of the vibration signal.

Presented herein are methods and systems for measuring the value of one or more parameters of interest, including changes in the value of one or more parameters of interest, based on measured spectral differences. A trained spectral difference-based measurement model determines changes in the value of one or more parameters of interest based on spectral difference measurements measured before and after one or more process steps. In some examples, the spectral difference measurements are determined based on differences in measured intensity, differences in harmonic signal values, or differences in the values ​​of one or more Mueller matrix elements. The spectral difference measurements can be expressed as difference values, scalar values, or a set of coefficients of a function that fits the difference values. The spectral difference measurements can be determined based on weighting the spectral differences according to wavelength.

A matrix multiply engine can include a first operand buffer and a second operand buffer, each of which can store multiple operand elements arranged in rows and columns. A cell array can be formed of cells, where each cell includes a memory and accumulator circuitry to receive operand elements column-wise from each of the first operand buffer and the second operand buffer, to compute a dot product of the received operand elements, and to accumulate the dot product into a corresponding tile state element in the memory. Matrix elements of the operand matrices to be multiplied can be loaded row-wise into rows of the operand buffers and read column-wise into the cells. The number of elements for which a dot product is computed can be selected depending on operand element width.

One embodiment provides a graphics processor comprising an instruction cache to store an instruction and a compute block configured to perform multiply-accumulate operations in response to execution of the instruction. The compute block includes a scheduler to schedule a plurality of threads for execution of the instruction and multiply-accumulate circuitry configured to execute the instruction via the plurality of threads, wherein the multiply-accumulate circuitry includes a plurality of functional units configured to process, in parallel via the plurality of threads, a corresponding plurality of matrix elements to multiply a first matrix and a second matrix, and to multiply the first matrix and the second matrix includes to multiply data elements in a row of the first matrix by corresponding data elements in a column of the second matrix to generate a plurality of products.

The invention relates to an iron-nickel-based medium-temperature electrothermal alloy and a preparation method thereof, and belongs to the technical field of alloy preparation. The electrothermal alloy comprises the following chemical components in percentage by mass: 0.01%-0.03% of C, 0.70%-1.50% of Al, less than or equal to 0.80% of Cr, less than or equal to 0.003% of S, less than or equal to 0.020% of P, 51%-53% of Ni, 0.04%-0.20% of rare earth, 0.80%-1.50% of Ti and a matrix element Fe, wherein the rare earth comprises at least one of Y, La and Ce. Alloy elements such as Al, rare earth and Ti are introduced, so that the oxidation resistance of the alloy is improved, and the service life of the alloy in a medium-temperature environment is greatly prolonged. And meanwhile, on the basis of ensuring excellent mechanical properties, the resistivity of the alloy material is effectively improved, and the resistance temperature coefficient of the alloy material is optimized, so that the alloy material can better meet the performance requirement of the positive temperature coefficient effect, and finally, the applicability, design flexibility and integration degree of the electrothermal alloy in a limited design space are enhanced.

The invention discloses a nickel-based corrosion-resistant alloy and a processing method thereof, and belongs to the technical field of nickel-based corrosion-resistant alloys, and the nickel-based corrosion-resistant alloy comprises the following chemical components in percentage by weight: 0.01-0.018% of carbon, 0.07-0.15% of silicon, 0.08-0.15% of manganese, 0.007-0.015% of phosphorus, less than or equal to 0.005% of sulfur, 26.5-27.5% of chromium, 8.5-9.6% of molybdenum, 0.22-0.37% of aluminum, 3.2-3.8% of niobium, 5.3-6.5% of iron, 0.01-0.018% of nitrogen and less than or equal to 0.002% of oxygen. The trace impurities are inevitably introduced in the production process, the trace impurities are derived from raw material introduction and environmental contact in smelting and processing links, the total content of the trace impurities is smaller than or equal to 0.5%, and the content of a single impurity element is smaller than or equal to 0.05%. According to the mode, by reasonably adjusting the element ratio, on the basis that it is ensured that nickel serves as a matrix element to form a stable austenite structure (the cold and hot machining performance and the use performance of the alloy are stable), the nickel content is properly reduced, the use amount of precious metal elements is reduced, and therefore the alloy cost is reduced on the raw material level; the problem that wide application of existing alloy is limited due to the high proportion of precious metal is solved.

The invention relates to 420 MPa grade alkaline soil corrosion resistant steel and a preparation method, and belongs to the technical field of steel preparation. The alkali soil corrosion resistant steel comprises the following chemical components in percentage by mass: 0.03%-0.12% of C, 0.2%-0.6% of Si, 0.5%-2.0% of Mn, 0.6%-3.5% of Cr, 0.6%-1.2% of Al, 0.001%-0.3% of Mo, 0.001%-0.5% of Ni, 0.001%-0.5% of Cu, less than or equal to 0.008% of P, less than or equal to 0.006% of S, less than or equal to 0.1% of Sn, less than or equal to 0.1% of Sb and a matrix element Fe. Specific chemical components including C, Si, Mn, Cr, Al, Mo, Ni, Cu and other alloy elements are designed, the content range of all the elements is strictly controlled, through the composite effect of all the elements, the steel can form a compact oxidation film in a corrosion environment, corrosion of a corrosion medium is effectively resisted, and therefore the corrosion resistance of the steel is improved, the service life of the steel is prolonged, and the service life of the steel is prolonged. And a powerful guarantee is provided for long-term application of the steel structure in the northwest alkaline soil environment, and wide application prospects are achieved.

The invention provides a matrix distribution and calculation method and system for multiply-add units, and a storage medium. The method comprises the following steps: determining the number of rows and columns of an array formed by a plurality of multiply-add units; a matrix needing matrix operation is obtained by inputting a request source, and the number of rows and columns of matrix elements obtained through operation is determined in advance; determining the number of multiplication and addition units needing to be occupied according to the row and column number of the calculated matrix elements, and endowing the multiplication and addition units with corresponding row and column numbers and occupation indexes; the first matrix elements participating in the operation are sequentially distributed to the multiplication and addition units of the corresponding column numbers according to columns, and the second matrix elements participating in the operation are sequentially distributed to the multiplication and addition units of the corresponding row numbers according to rows; based on a multiply-add unit, matrix calculation is carried out according to the distributed matrix elements, and an intermediate result is obtained; and through an output request, taking out intermediate results from the multiply-add unit and sequentially splicing the intermediate results to obtain a matrix operation result. According to the invention, the resource utilization rate of the multiply-add unit array can be improved.

The invention discloses a resource allocation method and device, a computer program product and electronic equipment. The method comprises the following steps: acquiring project information of a target project; multiple evaluation indexes of the target project are determined according to the project information, index weights of the evaluation indexes are determined based on an index judgment matrix, and matrix elements of the index judgment matrix are used for reflecting the importance proportion between every two evaluation indexes; the feasibility level of the target project is determined based on a fuzzy relation matrix and the index weight of each evaluation index, and matrix elements of the fuzzy relation matrix are used for representing the membership degree of each evaluation index to comments of different levels; and determining a cloud resource allocation strategy of the target project according to the project information and the feasibility level, and allocating corresponding cloud resources to the target project according to the cloud resource allocation strategy. The technical problems of unreasonable resource allocation and untimely risk response existing in a cloud resource scheduling strategy for formulating a project based on static information are solved.