DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Hwang, Ganguk | - |
dc.contributor.advisor | 황강욱 | - |
dc.contributor.author | Kim, Jeongseop | - |
dc.date.accessioned | 2021-05-11T19:43:49Z | - |
dc.date.available | 2021-05-11T19:43:49Z | - |
dc.date.issued | 2020 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=907854&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/283583 | - |
dc.description | 학위논문(박사) - 한국과학기술원 : 수리과학과, 2020.2,[iv, 59 p. :] | - |
dc.description.abstract | Next generation networks with extremely high data rates and various applications require intelligent networking systems that can support such demands reliably. Recently, machine learning paradigm has attracted increasing attention as a promising technology for such systems. In this dissertation we consider Gaussian process regression (GPR) based intelligent networking systems that guarantee given quality of service (QoS) requirements. In Chapter 3, we propose an adaptive bandwidth allocation algorithm to satisfy the overflow probability QoS. In the proposed algorithm with GPR we consider the stochastic property of each sample path individually and compute the required bandwidth adaptively for each sample path by estimating its overflow probability. We investigate the computational complexity and performance of the proposed algorithm through simulations with real-world traffic as well as synthetic traffic and show that the proposed algorithm allocates the bandwidth adaptively and efficiently to satisfy the required QoS. We also analyze large-buffer asymptotics for Gaussian queues including non-stationary input processes to further improve the resource efficiency of the proposed algorithm in Chapter 4. Next, we consider traffic load balancing in a multimedia multipath (MMMP) system that utilizes a variety of communication media and paths. In Chapter 5, we propose an online load balancing algorithm which does not require any information about the system. To this end, we introduce a probing period to collect training data for GPR and estimate the timeout probability of each path using the predictive distributions from GPR. We then analyze and minimize the cost function, the weighted sum of the timeout probabilities of the paths. Through extensive simulations under various scenarios, we demonstrate that the proposed algorithm balances the traffic load properly according to dynamic system conditions. Furthermore, we analyze the impact of probing packets on the network and design the proposed algorithm to reduce the communication overhead due to probing packets. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Gaussian process regression▼atraffic prediction▼aadaptive bandwidth allocation▼aload balancing▼amultimedia multipath system▼aonline learning | - |
dc.subject | 가우시안 확률과정 회귀 모델▼a트래픽 예측▼a적응적 대역폭 할당▼a부하 분산▼a다매체 다중경로 시스템▼a온라인 학습 | - |
dc.title | Intelligent network traffic analysis based on gaussian process regression | - |
dc.title.alternative | 가우시안 확률과정 회귀 모델을 이용한 지능형 네트워크 트래픽 분석 | - |
dc.type | Thesis(Ph.D) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :수리과학과, | - |
dc.contributor.alternativeauthor | 김정섭 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.