公开(公告)号：US20250045054A1

公开(公告)日：2025-02-06

申请号：US18781624

申请日：2024-07-23

Applicant: SAMSUNG ELECTRONICS CO., LTD. ,  POSTECH Research and Business Development Foundation

Inventor： Jin HYUN ,  Youngjoo LEE ,  Seungwoo HONG

IPC: G06F9/30

Abstract: A processor-implemented method includes determining a first width, the first width being a minimum width of a plurality of sectors into which an input range for a non-linear function is divided, and a plurality of second widths dividing the plurality of sectors each into one or more segments, determining a final width of a target sector, the target sector being one of the plurality of sectors, for approximating the non-linear function, based on one or more among the first width and the plurality of second widths, and dividing the non-linear function into one or more segments comprised in the target sector and approximating the divided non-linear function for each of the one or more segments to a linear function.

公开(公告)号：US20240178938A1

公开(公告)日：2024-05-30

申请号：US18458945

申请日：2023-08-30

Applicant: POSTECH Research and Business Development Foundation

Inventor： Youngjoo LEE ,  Young Seok KIM ,  Dain PARK

IPC: H04L1/00

CPC classification number: H04L1/0063 ,  H04L1/0052

Abstract: A polar code-based decoding method may comprise: converting a plurality of LLRs of calculation-target bits into a plurality of absolute (ABS) LLR values; sorting the plurality of ABS LLR values in order of size so that a minimum ABS LLR value is designated from among the plurality of ABS LLR values; multiplying the minimum ABS LLR value by a preset weight; and updating path metrics (PMs) of the calculation-target bits in an order in which the plurality of ABS LLR values are sorted, wherein the updating of the PMs of the calculation-target bits comprises, even when a specific ABS LLR value of a specific calculation-target bit is equal or similar to the minimum ABS LLR value, updating a PM of the specific calculation-target bit with the minimum ABS LLR value multiplied by the preset weight.

公开(公告)号：US20240176654A1

公开(公告)日：2024-05-30

申请号：US18317004

申请日：2023-05-12

Applicant: POSTECH RESEARCH AND BUSINESS DEVELOPMENT FOUNDATION

Inventor： Youngjoo LEE ,  Hyeokjun KWON ,  Young-Seok KIM

IPC: G06F9/48 ,  G06F5/01 ,  G06F7/50 ,  G06F7/523 ,  G06F7/544 ,  G06F9/50

CPC classification number: G06F9/4881 ,  G06F5/01 ,  G06F7/50 ,  G06F7/523 ,  G06F7/5443 ,  G06F9/5027

Abstract: Provided is a method of optimizing calculation of a convolutional layer including a plurality of channels and a convolutional neural network (CNN) accelerator. The method includes a classification operation of classifying weights of the plurality of channels as outlier (OL) elements and non-outlier (NOL) elements, a combination operation of combining two or more NOL elements which are put into calculation with the same activation value and present at the same slot in the plurality of channels, and a scheduling operation of moving a scheduling candidate element in a fetch window after the combination operation to a slot which is assigned no value.