Resistive processing units and neural network training methods

    公开(公告)号:US10664745B2

    公开(公告)日:2020-05-26

    申请号:US15196346

    申请日:2016-06-29

    Abstract: An array of resistive processing units (RPUs) comprises a plurality of rows of RPUs and a plurality of columns of RPUs wherein each RPU comprises an AND gate configured to perform an AND operation of a first stochastic bit stream received from a first stochastic translator translating a number encoded from a neuron in a row and a second stochastic bit stream received from a second stochastic translator translating a number encoded from a neuron in a column. A first storage is configured to store a weight value of the RPU, and a second storage is configured to store an amount of change to the weight value of the RPU. When the first stochastic bit stream and the second stochastic bit stream coincide, the amount of change to the weight value of the RPU is added to the weight value of the RPU.

    RESISTIVE PROCESSING UNITS AND NEURAL NETWORK TRAINING METHODS

    公开(公告)号:US20180005110A1

    公开(公告)日:2018-01-04

    申请号:US15196346

    申请日:2016-06-29

    CPC classification number: G06N3/063 G06N3/0472 G06N3/084

    Abstract: An array of resistive processing units (RPUs) comprises a plurality of rows of RPUs and a plurality of columns of RPUs wherein each RPU comprises an AND gate configured to perform an AND operation of a first stochastic bit stream received from a first stochastic translator translating a number encoded from a neuron in a row and a second stochastic bit stream received from a second stochastic translator translating a number encoded from a neuron in a column. A first storage is configured to store a weight value of the RPU, and a second storage is configured to store an amount of change to the weight value of the RPU. When the first stochastic bit stream and the second stochastic bit stream coincide, the amount of change to the weight value of the RPU is added to the weight value of the RPU.

    AUTOMATED DECISION SUPPORT PROVENANCE AND SIMULATION

    公开(公告)号:US20160283849A1

    公开(公告)日:2016-09-29

    申请号:US14742882

    申请日:2015-06-18

    CPC classification number: G06N5/043 G06Q10/00 G06Q10/06 G06Q50/00

    Abstract: Embodiments relate to supporting a decision making process. The method generates a graph that represents a decision making process. The graph comprises a plurality of nodes and a plurality of edges connecting the nodes. The nodes represent local decisions contributing to a global decision of the decision making process. Each node is associated with one or more parameters used for modeling the local decision. Each edge is associated with one or more parameters used for defining a relationship between two nodes. The method simulates the graph based at least in part on the parameters of the nodes and edges to derive an output global decision of the decision making process. The method receives a change to at least one of the parameters of the graph from a user and simulates the graph based at least in part on the at least one changed parameter to determine that the output global decision changes.

    AUTOMATED DECISION SUPPORT PROVENANCE AND SIMULATION
    5.
    发明申请
    AUTOMATED DECISION SUPPORT PROVENANCE AND SIMULATION 有权
    自动化决策支持保证和模拟

    公开(公告)号:US20160283848A1

    公开(公告)日:2016-09-29

    申请号:US14666940

    申请日:2015-03-24

    CPC classification number: G06N5/043 G06Q10/00 G06Q10/06 G06Q50/00

    Abstract: Embodiments relate to supporting a decision making process. The method generates a graph that represents a decision making process. The graph comprises a plurality of nodes and a plurality of edges connecting the nodes. The nodes represent local decisions contributing to a global decision of the decision making process. Each node is associated with one or more parameters used for modeling the local decision. Each edge is associated with one or more parameters used for defining a relationship between two nodes. The method simulates the graph based at least in part on the parameters of the nodes and edges to derive an output global decision of the decision making process. The method receives a change to at least one of the parameters of the graph from a user and simulates the graph based at least in part on the at least one changed parameter to determine that the output global decision changes.

    Abstract translation: 实施例涉及支持决策过程。 该方法生成一个表示决策过程的图表。 该图包括连接节点的多个节点和多个边缘。 这些节点代表了决策过程的全球决策的本地决策。 每个节点与用于建模本地决策的一个或多个参数相关联。 每个边缘与用于定义两个节点之间的关系的一个或多个参数相关联。 该方法至少部分地基于节点和边缘的参数来模拟图,以得出决策过程的输出全局决策。 该方法从用户接收对图形的至少一个参数的改变,并且至少部分地基于至少一个改变的参数来模拟图,以确定输出全局决定改变。

    Automated decision support provenance and simulation

    公开(公告)号:US10169710B2

    公开(公告)日:2019-01-01

    申请号:US14742882

    申请日:2015-06-18

    Abstract: Embodiments relate to supporting a decision making process. The method generates a graph that represents a decision making process. The graph comprises a plurality of nodes and a plurality of edges connecting the nodes. The nodes represent local decisions contributing to a global decision of the decision making process. Each node is associated with one or more parameters used for modeling the local decision. Each edge is associated with one or more parameters used for defining a relationship between two nodes. The method simulates the graph based at least in part on the parameters of the nodes and edges to derive an output global decision of the decision making process. The method receives a change to at least one of the parameters of the graph from a user and simulates the graph based at least in part on the at least one changed parameter to determine that the output global decision changes.

    Automated decision support provenance and simulation

    公开(公告)号:US09836695B2

    公开(公告)日:2017-12-05

    申请号:US14666940

    申请日:2015-03-24

    CPC classification number: G06N5/043 G06Q10/00 G06Q10/06 G06Q50/00

    Abstract: Embodiments relate to supporting a decision making process. The method generates a graph that represents a decision making process. The graph comprises a plurality of nodes and a plurality of edges connecting the nodes. The nodes represent local decisions contributing to a global decision of the decision making process. Each node is associated with one or more parameters used for modeling the local decision. Each edge is associated with one or more parameters used for defining a relationship between two nodes. The method simulates the graph based at least in part on the parameters of the nodes and edges to derive an output global decision of the decision making process. The method receives a change to at least one of the parameters of the graph from a user and simulates the graph based at least in part on the at least one changed parameter to determine that the output global decision changes.

    METHOD, COMPUTER PROGRAM AND SYSTEM PROVIDING REAL-TIME POWER GRID HYPOTHESIS TESTING AND CONTIGENCY PLANNING
    8.
    发明申请
    METHOD, COMPUTER PROGRAM AND SYSTEM PROVIDING REAL-TIME POWER GRID HYPOTHESIS TESTING AND CONTIGENCY PLANNING 审中-公开
    方法,计算机程序和系统提供实时功率测量和连续性规划

    公开(公告)号:US20150006141A1

    公开(公告)日:2015-01-01

    申请号:US13968908

    申请日:2013-08-16

    Abstract: A data processing system includes a user interface with a user input configured to enable a user to specify a type of simulation to be performed and at least one initial condition, where the simulation is executed using at least one sensor input from a grid structure composed of at least one of a power transmission and distribution grid. The user interface further has a display configured to visualize a representation of a result of a simulation of at least one scenario by presenting a multi-dimensional representation comprised of indicators, where each indicator corresponds to at least one simulation result. The user interface responds to a selection of one of the indicators by the user to visualize a result of the corresponding simulation. The type of simulation can be an N−k contingency analysis simulation, where k is equal to zero, 1 or greater than 1.

    Abstract translation: 数据处理系统包括具有用户输入的用户界面,用户输入被配置为使得用户能够指定要执行的模拟的类型和至少一个初始状态,其中使用来自网格结构的至少一个传感器输入来执行模拟,所述网格结构由 至少一个输电和配电网。 用户接口还具有配置成通过呈现包括指示符的多维表示来可视化至少一个场景的模拟结果的表示的显示器,其中每个指示符对应于至少一个模拟结果。 用户界面响应用户选择一个指示符,以便可视化相应模拟的结果。 模拟类型可以是N-k应急分析模拟,其中k等于零,1或大于1。

    Minibatch Parallel Machine Learning System Design

    公开(公告)号:US20200050971A1

    公开(公告)日:2020-02-13

    申请号:US16058017

    申请日:2018-08-08

    Abstract: The disclosure is directed to optimizing parallel machine learning system design and performance using minibatch. A system for allocating data center resources according to embodiments includes: a machine learning process; a machine learning data set; a processing system including a P parallel processing elements for training the machine learning process using the machine learning data set, wherein the machine learning data set is split into a plurality of batches with a batch size M; and a resource manager for (1) minimizing a training time T=T(M,P) of the machine learning process over M for each value of P, and (2) efficient system design.

Patent Agency Ranking