-
公开(公告)号：US20210303972A1
公开(公告)日：2021-09-30
申请号：US17012181
申请日：2020-09-04
Applicant: Samsung Electronics Co., Ltd.
Inventor： Donghyun LEE , Sangwon HA
Abstract: A method for neural network quantization includes performing a forward pass and a backward pass of a first neural network having a first bit precision with respect to each of a plurality of input data sets, obtaining profile information with respect to at least one of input gradients, weight gradients, and output gradients calculated for each layer of layers included in the first neural network in the process of performing the backward pass, determining one or more layers, from among the layers, to be quantized with a second bit precision less than the first bit precision, based on the obtained profile information, and generating a second neural network by quantizing the determined layers from among the layers with the second bit precision.
-
公开(公告)号：US20210182594A1
公开(公告)日：2021-06-17
申请号：US16906326
申请日：2020-06-19
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sangwon HA
Abstract: A method of processing a neural network, includes generating an integral map for each channel in a first layer of the neural network based on calculating of area sums of pixel values in first output feature maps of channels in the first layer, generating an accumulated integral map by performing an accumulation operation on the integral maps generated for the respective channels, obtaining pre-output feature maps of a second layer, subsequent to the first layer, by performing a convolution operation between input feature maps of the second layer and weight kernels, and removing offsets in the weight kernels to obtain second output feature maps of the second layer by subtracting accumulated values of the accumulated integral map from pixel values of the pre-output feature maps.
-
公开(公告)号：US20230403795A1
公开(公告)日：2023-12-14
申请号：US18456153
申请日：2023-08-25
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sangwon HA , Subyung KANG , Sanghoon PARK , Soohyun SEO , Yongjin WOO , Yeomoon YOON
IPC: H05K1/14
CPC classification number: H05K1/145 , H05K1/144 , H01L2023/4087
Abstract: An electronic device is provided. The electronic device includes a housing, a first printed circuit board disposed in an internal space of the housing and including first conductive terminals, and a second printed circuit board disposed parallel to the first printed circuit board in the internal space and including second conductive terminals electrically connected to the first conductive terminals. The second printed circuit board includes at least some conductive terminals of the second conductive terminals, or at least one connection failure prevention structure disposed around at least some conductive terminals of the second conductive terminals.
-
公开(公告)号：US20210201117A1
公开(公告)日：2021-07-01
申请号：US16909095
申请日：2020-06-23
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sangwon HA , Gunhee KIM , Donghyun LEE
Abstract: A processor-implemented neural network method includes: determining a respective probability density function (PDF) of normalizing a statistical distribution of parameter values, for each channel of each of a plurality of feature maps of a pre-trained neural network; determining, for each channel, a corresponding first quantization range for performing quantization of corresponding parameter values, based on a quantization error and a quantization noise of the respective determined PDF; determining, for each channel, a corresponding second quantization range, based on a signal-to-quantization noise ratio (SQNR) of the respective determined PDF; correcting, for each channel, the corresponding first quantization range based on the corresponding second quantization range; and generating a quantized neural network, based on the corrected first quantization range corresponding for each channel.
-
公开(公告)号：US20230017432A1
公开(公告)日：2023-01-19
申请号：US17950342
申请日：2022-09-22
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sangwon HA , Gunhee KIM , Donghyun LEE
Abstract: A processor-implemented neural network method includes: determining a respective probability density function (PDF) of normalizing a statistical distribution of parameter values, for each channel of each of a plurality of feature maps of a pre-trained neural network; determining, for each channel, a corresponding first quantization range for performing quantization of corresponding parameter values, based on a quantization error and a quantization noise of the respective determined PDF; determining, for each channel, a corresponding second quantization range, based on a signal-to-quantization noise ratio (SQNR) of the respective determined PDF; correcting, for each channel, the corresponding first quantization range based on the corresponding second quantization range; and generating a quantized neural network, based on the corrected first quantization range corresponding for each channel.
-
Patent Agency Ranking
6.发明申请公开(公告)号：US20210091450A1
公开(公告)日：2021-03-25
申请号：US17025096
申请日：2020-09-18
Applicant: Samsung Electronics Co., Ltd.
Inventor： Jungmin PARK , Chonghwa SEO , Dongil SON , Jongwon LEE , Sangwon HA
Abstract: An electronic device is provided. The electronic device includes a housing, an antenna module disposed along an inner wall of the housing, and a heat dissipation member disposed inside the housing. The antenna module may include a first circuit board including a plurality of layers, wherein the plurality of layers include a plurality of conductive layers and a plurality of non-conductive layers, and the conductive layer and the non-conductive layer are alternately stacked, at least one integrated circuit mounted on the first circuit board, an antenna array disposed on any one of the plurality of layers and electrically coupled with the at least one integrated circuit, and a heat transfer member including a conductive member extending from any one conductive layer among the plurality of conductive layers and a non-conductive member surrounding, at least in part, the conductive member.
-
公开(公告)号：US20210014970A1
公开(公告)日：2021-01-14
申请号：US16883604
申请日：2020-05-26
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sangwon HA , Seyoung JANG , Sungjin KIM , Sanghoon PARK , Kyungho LEE , Younoh CHI
IPC: H05K1/14 , H01L25/065
Abstract: An electronic device is provided that includes a first circuit board including a first electronic component and a second electronic component disposed on a side of the first circuit board, a second circuit board spaced apart from the first circuit board and having a side facing the side of the first circuit board on which the first electronic component and the second electronic component are disposed, a first interposer disposed between the first circuit board and the second circuit board to form an inner space between the first circuit board and the second circuit board, and a second interposer disposed between the first circuit board and the second circuit board to divide the inner space into a first region and a second region, wherein the first electronic component on the first circuit board is disposed corresponding to the first region, and the second electronic component on the first circuit board is disposed corresponding to the second region, and wherein the first interposer and the second interposer electrically connect the first circuit board to the second circuit board.
-
公开(公告)号：US20230117033A1
公开(公告)日：2023-04-20
申请号：US18084948
申请日：2022-12-20
Applicant: Samsung Electronics Co., Ltd.
Inventor： Junhaeng LEE , Seungwon LEE , Sangwon HA , Wonjo LEE
Abstract: A method of generating a fixed-point quantized neural network includes analyzing a statistical distribution for each channel of floating-point parameter values of feature maps and a kernel for each channel from data of a pre-trained floating-point neural network, determining a fixed-point expression of each of the parameters for each channel statistically covering a distribution range of the floating-point parameter values based on the statistical distribution for each channel, determining fractional lengths of a bias and a weight for each channel among the parameters of the fixed-point expression for each channel based on a result of performing a convolution operation, and generating a fixed-point quantized neural network in which the bias and the weight for each channel have the determined fractional lengths.
-
公开(公告)号：US20220335705A1
公开(公告)日：2022-10-20
申请号：US17846202
申请日：2022-06-22
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sangwon HA
Abstract: A method of processing a neural network, includes generating an integral map for each channel in a first layer of the neural network based on calculating of area sums of pixel values in first output feature maps of channels in the first layer, generating an accumulated integral map by performing an accumulation operation on the integral maps generated for the respective channels, obtaining pre-output feature maps of a second layer, subsequent to the first layer, by performing a convolution operation between input feature maps of the second layer and weight kernels, and removing offsets in the weight kernels to obtain second output feature maps of the second layer by subtracting accumulated values of the accumulated integral map from pixel values of the pre-output feature maps.
-
公开(公告)号：US20210360796A1
公开(公告)日：2021-11-18
申请号：US17286519
申请日：2019-10-18
Applicant: Samsung Electronics Co., Ltd.
Inventor： Sungchul PARK , Sangwon HA , Chulwoo PARK
Abstract: A substrate connection member according to various embodiments of the present invention can comprise a printed circuit board which has a plurality of layers that are stacked and which comprises a front surface, a rear surface, and a side surface encompassing the front surface and the rear surface. The printed circuit board can comprise: an opening part which encompasses a partial region of the printed circuit board and which is penetratingly formed from the front surface to the rear surface; at least one bridge connected between the partial region and the printed circuit board by crossing at least a portion of the opening part; and at least one through-hole wire formed in the partial region from the front surface to the rear surface, wherein the inner surface of the opening part and the side surface of the bridge can be formed from a conductive member. Other various embodiments, in addition to the embodiments disclosed in the present invention, are possible.
-
-
-
-
-
-
-
-
-
Search Results
17
records
(took 0.021 seconds)
