ELECTRONIC APPARATUS AND CONTROL METHOD THEREOF

An electronic apparatus is disclosed to perform deep learning. The electronic apparatus includes: a storage configured to store target data and kernel data; and a processor including a plurality of processing elements that are arranged in a matrix shape. The processor is configured to input, to each...

Full description

Saved in:
Bibliographic Details
Main Authors PARK YOUNG HWAN, KIM DAE HYUN, CHO HAN SU, KIM HYUN JUNG, SUH DONG KWAN, PRASAD NAGARAJA KESHAVA, KIM SUK JIN, KIM KYOUNG HOON
Format Patent
LanguageEnglish
Korean
Published 22.04.2019
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:An electronic apparatus is disclosed to perform deep learning. The electronic apparatus includes: a storage configured to store target data and kernel data; and a processor including a plurality of processing elements that are arranged in a matrix shape. The processor is configured to input, to each of the plurality of processing elements, a first non-zero element from among a plurality of first elements included in the target data, and to input, to each of a plurality of first processing elements included in a first row from among the plurality of processing elements, a second non-zero element from among the plurality of elements included in the kernel data. Each of the plurality of first processing elements is configured to perform an operation between the input first non-zero element and the input second non-zero element, based on depth information of the first non-zero element and depth information of the second non-zero element. The present invention can skip calculation for portions of the target data and the kernel data according to the zero elements included in the target data to increase a convolution calculation speed. 딥러닝(Deep Learning)을 수행하는 전자 장치가 개시된다. 본 전자 장치는 대상 데이터 및 커널 데이터가 저장된 스토리지 및 매트릭스 형태로 배열된 복수의 연산 소자(Processing Element)를 포함하는 프로세서를 포함하고, 프로세서는 복수의 연산 소자 각각에 대상 데이터에 포함된 복수의 제1 엘리먼트 중 제1 논제로(non-zero) 엘리먼트를 입력하며, 커널 데이터에 포함된 복수의 제2 엘리먼트 중 제2 논제로 엘리먼트를 복수의 연산 소자 중 제1 로우에 포함된 복수의 제1 연산 소자 각각에 순차적으로 입력하고, 복수의 제1 연산 소자 각각은 입력된 제1 논제로 엘리먼트의 뎁스(depth) 정보 및 입력된 제2 논제로 엘리먼트의 뎁스 정보에 기초하여, 입력된 제1 논제로 엘리먼트 및 입력된 제2 논제로 엘리먼트 간 연산을 수행할 수 있다.
Bibliography:Application Number: KR20180022960