中科微点-全球专利检索

  1. Login
  2. Sign Up

Search Results

10 records (took 0.017 seconds)

    Feature extraction using multi-task learning
    1.
    发明授权
    Feature extraction using multi-task learning 有权

    公开(公告)号：US11100399B2

    公开(公告)日：2021-08-24

    申请号：US15818877

    申请日：2017-11-21

    Applicant: INTERNATIONAL BUSINESS MACHINES CORPORATION

    Inventor： Wei Shan Dong ,  Peng Gao ,  Chang Sheng Li ,  Chun Yang Ma ,  Kai AD Yang ,  Ren Jie Yao ,  Ting Yuan ,  Jun Zhu

    IPC: G06N3/08 ,  G06N3/04

    Abstract: Systems and methods for training a neural network model are disclosed. In the method, training data is obtained by a deep neural network (DNN) first, the deep neural network comprising at least one hidden layer. Then features of the training data are obtained from a specified hidden layer of the at least one hidden layer, the specified hidden layer being connected respectively to a supervised classification network for classification tasks and an autoencoder based reconstruction network for reconstruction tasks. And at last the DNN, the supervised classification network and the reconstruction network are trained as a whole based on the obtained features, the training being guided by the classification tasks and the reconstruction tasks.

    FEATURE EXTRACTION USING MULTI-TASK LEARNING
    4.
    发明申请
    FEATURE EXTRACTION USING MULTI-TASK LEARNING 审中-公开

    公开(公告)号：US20190156211A1

    公开(公告)日：2019-05-23

    申请号：US15818877

    申请日：2017-11-21

    Applicant: INTERNATIONAL BUSINESS MACHINES CORPORATION

    Inventor： Wei Shan Dong ,  Peng Gao ,  Chang Sheng Li ,  Chun Yang Ma ,  Kai AD Yang ,  Ren Jie Yao ,  Ting Yuan ,  Jun Zhu

    IPC: G06N3/08 ,  G06N3/04

    Abstract: Systems and methods training a model are disclosed. In the method, training data is obtained by a deep neural network (DNN) first, the deep neural network comprising at least one hidden layer. Then features of the training data are obtained from a specified hidden layer of the at least one hidden layer, the specified hidden layer being connected respectively to a supervised classification network for classification tasks and an autoencoder based reconstruction network for reconstruction tasks. And at last the DNN, the supervised classification network and the reconstruction network are trained as a whole based on the obtained features, the training being guided by the classification tasks and the reconstruction tasks

    Predicting a vehicular route of travel without historical route data
    8.
    发明授权
    Predicting a vehicular route of travel without historical route data 有权

    公开(公告)号：US10203218B2

    公开(公告)日：2019-02-12

    申请号：US15444936

    申请日：2017-02-28

    Applicant: INTERNATIONAL BUSINESS MACHINES CORPORATION

    Inventor： Wei Shan Dong ,  Ning Duan ,  Guo Qiang Hu ,  Ting Yuan ,  Jun Zhu

    IPC: G01C21/00 ,  G01C21/36 ,  G01S19/42 ,  G01C21/34

    Abstract: A method according to the present invention includes predicting a vehicular route. GPS data of a vehicle's position on a road network is received. A digital map representing the road network is received. The digital map includes a plurality of partitioned regions. Each of the partitioned regions includes a plurality of geographic nodes. A starting node is selected. At least one partitioned region is selected based on a predetermined travel-time horizon of the vehicle from the starting node. Route simulation is performed between the plurality of geographic nodes of the selected at least one partitioned region and a plurality of potential future routes is generated. An actual route of the vehicle is detected. The actual route of the vehicle is compared with the plurality of potential future routes. A probability of the vehicle traveling along each potential future route is determined. A future route of the vehicle is predicted.

    Method of effective driving behavior extraction using deep learning
    9.
    发明授权
    Method of effective driving behavior extraction using deep learning 有权

    公开(公告)号：US10198693B2

    公开(公告)日：2019-02-05

    申请号：US15332407

    申请日：2016-10-24

    Applicant: International Business Machines Corporation

    Inventor： Wei Shan Dong ,  Peng Gao ,  Jian Li ,  Chang Sheng Li ,  Wen Han Luo ,  Chun Yang Ma ,  Renjie Yao ,  Ting Yuan ,  Jun Zhu

    IPC: G06N3/08 ,  B60W40/08 ,  B60W50/06 ,  G01S19/42 ,  G01S19/14 ,  G06N3/04

    Abstract: Systems and methods for obtaining vehicle operational data and driving context data from one or more monitoring systems, including converting the obtained vehicle operational data and driving context data into sequential vehicle operational feature data and sequential driving context feature data, calibrating the sequential vehicle operational feature data and the sequential driving context feature data temporally to form calibrated sequential vehicle operational feature data and calibrated sequential driving context feature data, constructing a sequence table of temporal sample points based on the calibrated sequential vehicle operational feature data and the calibrated sequential driving context feature data, feeding the sequence table into a deep neural network model for applying network learning to form a trained deep neural network model, extracting driving behavior features from the trained deep neural network model and analyzing the extracted driving behavior features to determine driving behavior characteristics of the driver.

Patent Agency Ranking