ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Conference Paper 임베디드 시스템에서의 기계학습을 위한 양자화 학습 방정식에 관한 연구
Cited - time in scopus Share share facebook twitter linkedin kakaostory
Authors
석진욱, 김정시
Issue Date
2019-11
Citation
대한전자공학회 학술 대회 (추계) 2019, pp.675-678
Publisher
대한전자공학회
Language
Korean
Type
Conference Paper
Abstract
We propose that a quantized learning equation for machine learning on an embedded system, in this paper. The proposed learning scheme can minimize quantization errors and is suitable for embedded systems when quantization is composed of integer or fixed-point fraction values applied to an optimization algorithm. Moreover, by the proposed methodology, it is possible to implement a machine learning algorithm capable of exhibiting sufficient optimization performance even in low-performance hardware. The simulation results show that the optimization solver based on the proposed quantized method provides sufficient performance.
KSP Keywords
Embedded system, Fixed-point, Machine Learning Algorithms, Optimization algorithm, Optimization performance, Quantization error, simulation results