中科微点-全球专利检索

  1. Login
  2. Sign Up

Search Results

5 records (took 0.02 seconds)

    NON-UNIFORM QUANTIZATION OF PRE-TRAINED DEEP NEURAL NETWORK
    1.
    发明申请
    NON-UNIFORM QUANTIZATION OF PRE-TRAINED DEEP NEURAL NETWORK 审中-公开

    公开(公告)号：US20200097823A1

    公开(公告)日：2020-03-26

    申请号：US16181326

    申请日：2018-11-05

    Applicant: Samsung Electronics Co., Ltd.

    Inventor： Hui CHEN ,  Ilia OVSIANNIKOV

    IPC: G06N3/08 ,  G06N3/04

    Abstract: A system and a method of quantizing a pre-trained neural network, includes determining by a layer/channel bit-width determiner for each layer or channel of the pre-trained neural network a minimum quantization noise for the layer or the channel for each master bit-width value in a predetermined set of master bit-width values; and selecting by a bit-width selector for the layer or the channel the master bit-width value having the minimum quantization noise for the layer or the channel. In one embodiment, the minimum quantization noise for the layer or the channel is based on a square of a range of weights for the layer or the channel that is multiplied by a constant to a negative power of a current master bit-width value.

    METHOD PERFORMED BY BASE STATION, BASE STATION AND COMPUTER READABLE STORAGE MEDIUM
    2.
    发明公开
    METHOD PERFORMED BY BASE STATION, BASE STATION AND COMPUTER READABLE STORAGE MEDIUM 审中-公开

    公开(公告)号：US20240056861A1

    公开(公告)日：2024-02-15

    申请号：US18329840

    申请日：2023-06-06

    Applicant: Samsung Electronics Co., Ltd.

    Inventor： Weili CUI ,  Meifang JING ,  Ying LI ,  Hui CHEN ,  Jiajia WANG ,  Fang XIE

    IPC: H04W24/10 ,  H04W16/28

    CPC classification number: H04W24/10 ,  H04W16/28

    Abstract: Embodiments of the disclosure provide a method performed by a base station, a base station and a computer readable storage medium. The method includes: instructing a user equipment (UE) to perform a beam measurement at a first beam level; determining, based on at least one of information on transmission capacity, mobility information and traffic information of the UE, whether to instruct the UE to perform a beam measurement at a second beam level; and receiving beam measurement results of the UE and performing beam scheduling, and wherein, a scheduled beam includes a beam at the first beam level or a beam at the second beam level; and serving cells of the base station are covered by each of the beam levels, and beams at different beam levels have different attributes. Part of the the implementation process of the scheme can be achieved by artificial intelligence. The disclosure saves measurement overhead while ensuring communication quality, and achieves the effects of improving cell throughput and user experience.

    NON-UNIFORM QUANTIZATION OF PRE-TRAINED DEEP NEURAL NETWORK
    3.
    发明申请
    NON-UNIFORM QUANTIZATION OF PRE-TRAINED DEEP NEURAL NETWORK 有权

    公开(公告)号：US20220269945A1

    公开(公告)日：2022-08-25

    申请号：US17744601

    申请日：2022-05-13

    Applicant: Samsung Electronics Co., Ltd.

    Inventor： Hui CHEN ,  Ilia OVSIANNIKOV

    IPC: G06N3/08 ,  G06N3/04

    Abstract: A system and a method of quantizing a pre-trained neural network, includes determining by a layer/channel bit-width determiner for each layer or channel of the pre-trained neural network a minimum quantization noise for the layer or the channel for each master bit-width value in a predetermined set of master bit-width values; and selecting by a bit-width selector for the layer or the channel the master bit-width value having the minimum quantization noise for the layer or the channel. In one embodiment, the minimum quantization noise for the layer or the channel is based on a square of a range of weights for the layer or the channel that is multiplied by a constant to a negative power of a current master bit-width value.

Patent Agency Ranking