利索-全球专利检索

  1. 登录
  2. 注册

搜索结果

41 条记录 (耗时 0.028 秒)

    DYNAMICALLY TUNING HYPERPARAMETERS DURING ML MODEL TRAINING
    2.
    发明公开
    DYNAMICALLY TUNING HYPERPARAMETERS DURING ML MODEL TRAINING 审中-公开

    公开(公告)号：US20230259813A1

    公开(公告)日：2023-08-17

    申请号：US17674808

    申请日：2022-02-17

    申请人： INTERNATIONAL BUSINESS MACHINES CORPORATION

    发明人： Yuan-Chi Chang ,  Venkata Nagaraju Pavuluri ,  Dharmashankar Subramanian ,  Timothy Rea Dinger

    IPC分类号： G06N20/00 ,  G06N3/08 ,  G06N3/04

    CPC分类号： G06N20/00 ,  G06N3/08 ,  G06N3/0454

    摘要： A method of automatically tuning hyperparameters includes receiving a hyperparameter tuning strategy. Upon determining that one or more computing resources exceed their corresponding predetermined quota, the hyperparameter tuning strategy is rejected. Upon determining that the one or more computing resources do not exceed their corresponding predetermined quota, a machine learning model training is run with a hyperparameter point. Upon determining that one or more predetermined computing resource usage limits are exceeded for the hyperparameter point, the running of the machine learning model training is terminated for the hyperparameter point and the process returns to running the machine learning model training with a new hyperparameter point. Upon determining that training the machine learning model is complete, training results are collected and computing resource utilization metrics are determined. A correlation of the hyperparameters to the computing resource utilization is determined from the completed training of the machine learning model.

    Dual interactive visualization system for sensitivity analysis to risk preferences for decision support
    7.
    发明授权
    Dual interactive visualization system for sensitivity analysis to risk preferences for decision support 有权

    公开(公告)号：US10713303B2

    公开(公告)日：2020-07-14

    申请号：US14991117

    申请日：2016-01-08

    申请人： International Business Machines Corporation

    发明人： Maryam Ashoori ,  Debarun Bhattacharjya ,  Jeffrey O. Kephart ,  Dharmashankar Subramanian

    IPC分类号： G06F16/904

    摘要： A system, computer program product, and method is described to provide a visualization tool which portrays the certain equivalent for one or more hypothetical (i.e. synthetic) or real probability distributions p(m), and optionally allows the user to manipulate that representation, resulting in associated changes to the underlying utility function u(m). In a first example, the risk preference visualization tool allows one to explore how the certain equivalent depends upon the probability distribution p(m), for a fixed utility function u(m). In a second example, the risk preference visualization tool allows one to explore how the certain equivalent depends upon the utility function u(m), assuming one or more fixed probability distributions p1(m), p2 (m), etc. In a third example, the decision maker can provide feedback through the visualization tool that would cause their utility function to be modified.

    DATA-DRIVEN INVENTORY AND REVENUE OPTIMIZATION FOR UNCERTAIN DEMAND DRIVEN BY MULTIPLE FACTORS
    9.
    发明申请
    DATA-DRIVEN INVENTORY AND REVENUE OPTIMIZATION FOR UNCERTAIN DEMAND DRIVEN BY MULTIPLE FACTORS 审中-公开
    数据驱动的存货和收益优化由多个因素驱动的不确定性需求

    公开(公告)号：US20140365276A1

    公开(公告)日：2014-12-11

    申请号：US14033497

    申请日：2013-09-22

    申请人： International Business Machines Corporation

    发明人： Pavithra Harsha ,  Ramesh Natarajan ,  Dharmashankar Subramanian

    IPC分类号： G06Q10/06 ,  G06Q30/02

    CPC分类号： G06Q10/06315 ,  G06Q30/0202

    摘要： Based on a time series history of a random variable representing demand for at least one of a good and a service as a function of at least one controllable demand driver, obtain a quantile regression function that estimates a quantile of a demand distribution function; obtain a mixed- and/or super-quantile regression function that estimates conditional value at risk; and obtain a regression function that estimates mean of the demand distribution function. Joint optimization of: inventory of the at least one of a good and a service, and the at least one controllable demand driver, is undertaken based on the quantile regression function and the mixed- and/or super-quantile regression function, to obtain an optimal value for the at least one controllable demand driver and an implied optimal value for a stocking level. One or more exogenous demand drivers can optionally be taken into account.

    摘要翻译： 基于作为至少一个可控需求驱动器的函数的对于至少一个商品和服务的需求的随机变量的时间序列历史，获得估计需求分配函数的分位数的分位数回归函数; 获得估计风险条件值的混合和/或超分位数回归函数; 并获得估计需求分布函数的平均值的回归函数。 基于分位数回归函数和混合和/或超分位数回归函数进行联合优化：优点和服务中的至少一个以及至少一个可控需求驱动器的库存，以获得 对于至少一个可控需求驱动器的最佳值和放养级别的隐含最佳值。 可以可以考虑一个或多个外部需求驱动器。