中科微点-全球专利检索

  1. Login
  2. Sign Up

Search Results

53 records (took 0.029 seconds)

    METHOD AND DEVICE FOR NEURAL ARCHITECTURE SEARCH OPTIMIZED FOR BINARY NEURAL NETWORK
    41.
    发明申请
    METHOD AND DEVICE FOR NEURAL ARCHITECTURE SEARCH OPTIMIZED FOR BINARY NEURAL NETWORK 有权

    公开(公告)号：US20210264240A1

    公开(公告)日：2021-08-26

    申请号：US17105988

    申请日：2020-11-27

    Applicant: GIST(Gwangju Institute of Science and Technology)

    Inventor： Kunal Pratap SINGH ,  Da Hyun KIM ,  Jong Hyun CHOI

    IPC: G06N3/063 ,  G06N3/08 ,  G06N3/04 ,  G06N7/00

    Abstract: A method for generating a target network by performing neural architecture search using optimized search space is provided. The method includes steps of: a computing device (a) if a target data is inputted into the target network, allowing the target network to apply neural network operation to the target data, to generate an estimated search vector; and (b) allowing a loss layer to calculate architecture parameter losses by referring to the estimated search vector and a ground truth search vector, and to perform backpropagation by referring to the architecture parameter losses to update architecture parameter vectors for determining final layer operations among candidate layer operations included in an optimized layer type set corresponding to the optimized search space and wherein the final layer operations are to be performed by neural blocks, within cells of the target network, arranged according to an optimized cell template corresponding to the optimized search space.

    METHOD FOR GENERATING PERSONALIZED HRTF
    46.
    发明公开
    METHOD FOR GENERATING PERSONALIZED HRTF 审中-公开

    公开(公告)号：US20240354565A1

    公开(公告)日：2024-10-24

    申请号：US18434417

    申请日：2024-02-06

    Applicant: GIST(Gwangju Institute of Science and Technology)

    Inventor： Hong Kook KIM ,  Geon Woo LEE

    IPC: G06N3/08

    CPC classification number: G06N3/08

    Abstract: The present disclosure relates to a method for generating a personalized HRTF using a neural network model having a one-to-many structure. A method for generating a personalized HRTF according to an embodiment of the present disclosure includes: training a neural network model using multi-angle Head-Related Transfer Functions (HRTF) labeled to body information of a learning object; and obtaining multi-angle HRTFs at a time by inputting body information of a target user into the trained neural network model.

    Blood diagnostic device
    47.
    发明授权
    Blood diagnostic device 有权

    公开(公告)号：US12011718B2

    公开(公告)日：2024-06-18

    申请号：US17162753

    申请日：2021-01-29

    Applicant: GIST(Gwangju Institute of Science and Technology)

    Inventor： Sung Yang ,  Ji Chul Hyun ,  Yu Gyung Jung

    IPC: B01L3/00 ,  G01N15/10 ,  G01N15/1031 ,  G01N33/49 ,  G01N15/01

    CPC classification number: B01L3/502761 ,  G01N15/1023 ,  G01N15/1031 ,  G01N33/491 ,  B01L2200/0652 ,  B01L2300/0636 ,  B01L2300/0645 ,  B01L2300/0819 ,  G01N2015/016

    Abstract: The present invention relates to a blood diagnostic device, which includes one or more blood input parts into which blood is injected, a deterministic lateral displacement separation part which communicates with the blood input part to form a blood flow path along which the blood flows and separates white blood cells from remaining blood components, a first microchannel which communicates with the deterministic lateral displacement separation part so that the white blood cells separated through the deterministic lateral displacement separation part flow therein, one or more second microchannels which communicate with the deterministic lateral displacement separation part so that the remaining blood components separated through the deterministic lateral displacement separation part flow therein, a first discharge part which communicates with the first microchannel so that the white blood cells flowing in the first microchannel are discharged therethrough, and a second discharge part which communicates with the second microchannel.

    EARLY STOPPING METHOD FOR NEURAL NETWORK USING UNLABELED DATA
    48.
    发明公开
    EARLY STOPPING METHOD FOR NEURAL NETWORK USING UNLABELED DATA 审中-公开

    公开(公告)号：US20240013060A1

    公开(公告)日：2024-01-11

    申请号：US18161461

    申请日：2023-01-30

    Applicant: GIST (Gwangju Institute of Science and Technology)

    Inventor： Hyunju Lee ,  HongSeok Choi ,  Dongha Choi

    IPC: G06N3/09

    CPC classification number: G06N3/09

    Abstract: An early stopping method for a neural network according to an embodiment of the present disclosure includes: dividing a labeled dataset into a training dataset and a validation dataset; creating a pretrained neural network by training a neural network using the training dataset and early stopping learning of the neural network using the validation dataset; and creating a target neural network for each epoch by training the target neural network using the entire labeled dataset, and early stopping learning of the target neural network on the basis of a similarity between output of the pretrained neural network on at least one of the labeled data and unlabeled data and output of the target neural network on the unlabeled data.

Patent Agency Ranking