-
公开(公告)号：US12072953B2
公开(公告)日：2024-08-27
申请号：US17349817
申请日：2021-06-16
发明人： Gaurav Chadha , Sam Idicula , Sandeep Agrawal , Nipun Agarwal
摘要： Techniques are described herein for performing efficient matrix multiplication in architectures with scratchpad memories or associative caches using asymmetric allocation of space for the different matrices. The system receives a left matrix and a right matrix. In an embodiment, the system allocates, in a scratchpad memory, asymmetric memory space for tiles for each of the two matrices as well as a dot product matrix. The system proceeds with then performing dot product matrix multiplication involving the tiles of the left and the right matrices, storing resulting dot product values in corresponding allocated dot product matrix tiles. The system then proceeds to write the stored dot product values from the scratchpad memory into main memory.
-
公开(公告)号：US20240126812A1
公开(公告)日：2024-04-18
申请号：US18373870
申请日：2023-09-27
IPC分类号： G06F16/901 , G06F17/12
CPC分类号： G06F16/9024 , G06F17/12
摘要： Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a graph model representing an environment being interacted with by an agent. In one aspect, one of the methods include: obtaining experience data; using the experience data to update a visitation count for each of one or more state-action pairs represented by the graph model; and at each of multiple environment exploration steps: computing a utility measure for each of the one or more state-action pairs represented by the graph model; determining, based on the utility measures, a sequence of one or more planned actions that have an information gain that satisfies a threshold; and controlling the agent to perform the sequence of one or more planned actions to cause the environment to transition from a state characterized by a last observation received after a last action in the experience data into a different state.
-
公开(公告)号：US11893531B2
公开(公告)日：2024-02-06
申请号：US17896757
申请日：2022-08-26
发明人： Devanand R , Narayan Nandeda , Tushar Shekhar
IPC分类号： G06Q10/00 , G06Q10/087 , G06F17/13 , G06Q10/04 , G06F17/12
CPC分类号： G06Q10/087 , G06F17/12 , G06F17/13 , G06Q10/04
摘要： A system and method are disclosed for solving a supply chain planning problem modeled as a linear programming (LP) problem. Embodiments include receiving an LP problem representing a supply chain planning problem for a supply chain network comprising material buffers and resource buffers, partitioning the supply chain network at a complicating node into at least two supply chains sharing the complicating node, formulating a decomposed subproblem for each of the supply chains, calculating an effective dual based, at least in part, on a mathematical difference of at least two dual values calculated by solving the functional-based decomposed subproblems, and generating a globally-optimal LP solution to the LP problem using subgradient descent with the effective dual.
-
4.发明公开公开(公告)号：US20240007298A1
公开(公告)日：2024-01-04
申请号：US17978147
申请日：2022-10-31
发明人： Kyung Ah SHIM
CPC分类号： H04L9/3247 , H04L9/0861 , G06F17/12 , G06F17/16 , H04L9/50
摘要： Disclosed is an electronic device for computing a solution to a linear system using a block matrix. The electronic device includes a memory configured to store data, and a processor configured to perform a plurality of operations by executing instructions. The electronic device computes the solution to the linear system without directly computing an inverse matrix R−1 of a block matrix R, using an inverse matrix of at least one of a plurality of sub-blocks matrices A, B, C, and D that constitute a block matrix R representing the linear system, and an inverse matrix of at least one of combinations of the plurality of sub-block matrices A, B, C, and D. According, the solution to the linear system can be efficiently computed.
-
公开(公告)号：US20230385368A1
公开(公告)日：2023-11-30
申请号：US18320279
申请日：2023-05-19
申请人： ROLLS-ROYCE plc
发明人： Bryan L LAPWORTH
摘要： A method, performed on at least one computing device, of determining solutions to a number of linear matrix equations satisfying A{right arrow over (x)}={right arrow over (b)}, where A is a n×n matrix, {right arrow over (x)} is a column vector with n entries, and {right arrow over (b)} is a column vector with n entries, is disclosed. The method comprises determining a linear combination of unitary matrices that is equivalent to the matrix A; based on the linear combination of unitary matrices, determining a column vector {right arrow over (x)} that satisfies the linear matrix equation; forming an updated matrix A based on the obtained column vector {right arrow over (x)}; forming an updated column vector {right arrow over (b)} based on the obtained column vector {right arrow over (x)}; updating the coefficients of the linear combination of unitary matrices based on the updated column vector {right arrow over (x)}; and based on the updated linear combination of unitary matrices, determining an updated column vector {right arrow over (x)} that satisfies the updated linear matrix equation.
-
公开(公告)号：US20230252276A1
公开(公告)日：2023-08-10
申请号：US17727650
申请日：2022-04-22
发明人： Hieu Van Tran
CPC分类号： G06N3/0635 , G06F17/12 , H03M1/142
摘要： Numerous examples are disclosed for performing calibration of various electrical parameters in a deep learning artificial neural network. In one example, a system comprises a digital-to-analog converter for receiving an input of k bits and generating a first analog output, a mapping scalar for converting the first analog output into a second analog output, and an analog-to-digital converter for generating an output of n bits from the second analog output, where n is a different value than k.
-
公开(公告)号：US11698945B2
公开(公告)日：2023-07-11
申请号：US16832056
申请日：2020-03-27
摘要： An analog computing system with coupled non-linear oscillators can solve complex combinatorial optimization problems using the weighted Ising model. The system is composed of a fully-connected LC oscillator network with low-cost electronic components and compatible with traditional integrated circuit technologies. Each LC oscillator, or node, in the network can be coupled to each other node in the array with a multiply and accumulate crossbar array or optical interconnects. When implemented with four nodes, the system performs with single-run ground state accuracies of 98% on randomized MAX-CUT problem sets with binary weights and 84% with five-bit weight resolutions. The four-node system can obtain solutions within five oscillator cycles with a time-to-solution that scales directly with oscillator frequency. A scaling analysis suggests that larger coupled oscillator networks may be used to solve computationally intensive problems faster and more efficiently than conventional algorithms.
-
8.发明公开公开(公告)号：US20230169142A1
公开(公告)日：2023-06-01
申请号：US17919432
申请日：2020-06-04
发明人： Junya HATTORI , Masaya ENDO , Yuko OMAGARI
IPC分类号： G06F17/12
CPC分类号： G06F17/12
摘要： An optimal solution calculation device for an optimization problem includes an initial condition generation unit for generating an executable initial solution and an equality constraint set with respect to an optimization problem, an optimization calculation unit for calculating a solution of a simultaneous linear equation generated from an evaluation function and for calculating an evaluated solution that is a solution to minimize or maximize the evaluation function, and an update unit. A convergence determination unit of the optimization calculation unit determines that an iterative solution has converged when a residual norm is equal to or less than a convergence determination threshold value, and outputs the converged iterative solution as the evaluated solution. The update unit determines the evaluated solution as an optimal solution when an update of the equality constraint set is determined to be unnecessary, and the convergence determination threshold value is the first threshold value.
-
公开(公告)号：US11604419B2
公开(公告)日：2023-03-14
申请号：US17241142
申请日：2021-04-27
发明人： Joannes Jitse Venselaar , Anagnostis Tsiatmas , Samee Ur Rehman , Paul Christiaan Hinnen , Jean-Pierre Agnes Henricus Marie Vaessen , Nicolas Mauricio Weiss , Gonzalo Roberto Sanguinetti , Thomai Zacharopoulou , Martijn Maria Zaal
摘要： Methods of determining information about a patterning process. In a method, measurement data from a metrology process applied to each of a plurality of metrology targets on a substrate is obtained. The measurement data for each metrology target includes at least a first contribution and a second contribution. The first contribution is from a parameter of interest of a patterning process used to form the metrology target. The second contribution is from an error in the metrology process. The method further includes using the obtained measurement data from all of the plurality of metrology targets to obtain information about an error in the metrology process, and using the obtained information about the error in the metrology process to extract a value of the parameter of interest for each metrology target.
-
公开(公告)号：US20230048097A1
公开(公告)日：2023-02-16
申请号：US17815978
申请日：2022-07-29
申请人： Dotphoton AG
摘要： A method for processing image data having noise and information, including: acquiring input raw image data having pixels of an image sensor used to take the image data, processing the input data, and outputting processed image-output data. The step of acquiring input data includes acquiring an input-noise model from the input data, and the step of processing the input raw image data includes a preprocessing operation and determining an output-noise model adapted to reflect noise in the output data, and producing output raw-image data consistent with the output-noise model, and the step of outputting the processed image data includes storing and/or transmitting the output raw image data and the output-noise model, which together form the output data, in a manner linking the output raw image data to the output-noise model, thereby allowing processing of the output data, as input data, such that the processing is adapted for pipeline processing.
-
-
-
-
-
-
-
-
-
搜索结果
558 条记录 (耗时 0.024 秒)
