계층구조의 머신러닝 알고리즘을 이용한 시스템 모델링 및 패턴인식

Abstract

Machine learning is defined to learn from experience(E) with respect to some class of tasks(T) and performance measure(P) if its performance at tasks in(T), as measured by(P), improves with experience(E). There are three types of machine learning: Supervised Learning, Unsupervised Learning, and Reinforcement Learning. Recently, machine learning has been applied as an important topic of the 4th industrial revolution and applied in many fields such as virtual measurement of manufacturing process, estimation of failure cause, autonomous vehicle, and recommended service algorithm. In this thesis, we propose three systems based on machine learning algorithm. Firstly, we propose a polynomial linear regression algorithm for temperature compensation of current transducers and verify the applicability of the algorithm through simulation in MATLAB environment. Secondly, a k-means clustering algorithm is applied to a demodulation system in order to realize a low-illuminated visible light communication system, and a transmission speed of about twice that of VPPM (Variable Pulse Position Modulation), which is a typical visible light communication modulation and demodulation technique, can be realized. Lastly, we propose a hierarchical neural network algorithm for analyzing spectrometric data, and confirmed that the proposed system has a higher accuracy than the logistic regression algorithm.