• Journal of KIISE:Information Networking
• /
• v.28 no.2
• /
• pp.242-250
• /
• 2001

This paper deals with the TCP bandwidth guarantee problem in a differentiated serviccs(Diffserv) network. The Diffserv assured s<:rvice differentiates packet drop probabilities to guarantee the promised bandwidth even under network congestion. However a token buffer marker fails to show adequate performance because TCI' generates packets according to the unique Tel' congestion control mechanism. We propose a marker that uses a data buffer as well as a token buffer. The marker with a data buffer works well with the assured service mechanism because it smooths Tel' traffic. We showed that the marker with a data buffer achieves the target throughput better than a marker with a token buffer only. We also showed that the optimal buffer size is proportional to reserved throughput and HTT.

• Journal of KIISE
• /
• v.42 no.12
• /
• pp.1561-1567
• /
• 2015

Abstract Semi-supervised learning is an area in machine learning that employs both labeled and unlabeled data in order to train a model and has the potential to improve prediction performance compared to supervised learning. Graph-based semi-supervised learning has recently come into focus with two phases: graph construction, which converts the input data into a graph, and label inference, which predicts the appropriate labels for unlabeled data using the constructed graph. The inference is based on the smoothness assumption feature of semi-supervised learning. In this study, we propose an enhanced label inference algorithm by incorporating the importance of each vertex. In addition, we prove the convergence of the suggested algorithm and verify its excellence.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.7
• /
• pp.1078-1090
• /
• 2022

The distribution patterns of battery usage time data per battery level are able to affect the performance of estimating battery remaining time in mobile devices. Outliers may mainly affect the estimation performance of statistical regression methods. In this paper, we propose a software framework that detects and processes outliers to improve the estimation performance of statistical regression methods. The proposed framework first detects outliers that degrade the estimation performance. The proposed framework replaces outliers with smoothed data. The difference between an outlier and its replaced data will be properly distributed into individual data. Finally, individual data are reinforced to improve the estimation performance. The numerical results obtained by experimenting the proposed framework confirmed that it yielded good performance of estimating battery remaining time.

• Journal of Internet Computing and Services
• /
• v.8 no.2
• /
• pp.105-115
• /
• 2007

The up-to-moment dataset is built by combining the past dataset and the recent dataset. The proposal is to compute association rules in real time. This study proposed the model, $EM_{past'}$ and algorithm that is sensitive to time. It can be utilized in real time by applying partitioned combination law after dividing the past dataset into(k-1). Also, we suggested $EM^{ES}_{past}$ applying the exponential smoothing method to $EM^p_{past'}$ When the association rules of $EM_{past'}\;EM^w_{past'\;and\;EM^{ES}_{past}$ were compared, The simulation results showed that $EM^{ES}_{past}$ is most accurate for testing dataset than $EM_{past}$ and $EM^w_{past}$ in huge dataset.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.53 no.3
• /
• pp.124-131
• /
• 2016

Baseline correction is very important due to influence on performance of spectral analysis in application of spectroscopy. Baseline is often estimated by parameter selection using visual inspection on analyte spectrum. It is a highly subjective procedure and can be tedious work especially with a large number of data. For these reasons, it is an objective and automatic procedure is necessary to select optimal parameter value for baseline correction. Asymmetrically reweighted penalized least squares (arPLS) based on penalized least squares was proposed for baseline correction in our previous study. The method uses a new weighting scheme based on the generalized logistic function. In this study, we present an automatic selection of optimal parameter for baseline correction using arPLS. The method computes fitness and smoothness values of fitted baseline within available range of parameters and then selects optimal parameter when the sum of normalized fitness and smoothness gets minimum. According to the experimental results using simulated data with varying baselines, sloping, curved and doubly curved baseline, and real Raman spectra, we confirmed that the proposed method can be effectively applied to optimal parameter selection for baseline correction using arPLS.

• Journal of the Korean Data and Information Science Society
• /
• v.4
• /
• pp.137-146
• /
• 1993

The Proposed goodness-of-fit test Statistic $\hat{\lambda}_{\alpha}$ derived from the test Statistc in Kim (1992) is itself a smoothing parameter which is selected to minimize an estimated MISE for a truncated series estimator, $d_{\hat{\lambda}{n}}$, of the comparison density function. Therefore, this test statistic leads immediately to a point estimate of the density function in the event that $H_{0}$ is ejected. The limiting distribution of $\hat{\lambda}_{\alpha}$ was obtained under the null hypothesis. It is also shown that this test is consistent against fixed alternatives.

• Journal of the Korean Data and Information Science Society
• /
• v.8 no.1
• /
• pp.9-20
• /
• 1997

Several methods are proposed for optimizing Fourier series estimators with respect to Mean Integrated Square Error metrics. Traditionally, such method have followed. one of two basic strategies; A stopping rules or the rules of determine multipliers. A central hypothesis of this study is that better estimates can be obtained by combining the two strategies. A new multiplier sequence is proposed, which used in conjunction with any of the stopping rules, is shown to improve the performance of estimator which relies solely on a stopping rule.

• Annual Conference on Human and Language Technology
• /
• 2005.10a
• /
• pp.134-138
• /
• 2005

본 논문은 한국어 TTS(Text-To-Speech)에서 운율 경계를 추정하는 문제를 클래스 분류문제로 보고 CRF(Conditional Random Fields)를 적용하여 운율 경계를 추정하였다. 우리는 품사와 운율 경계로 구성된 말뭉치를 사용하여 품사, 어휘, 단어의 길이, 문장에서의 단어 위치와 같은 다양한 속성의 언어적 자질을 추출하여 CRF를 훈련시켰으며, 자질들을 서로 조합하여 최고의 성능을 보이는 자질 집합을 골랐다 또한 가우스 평활 (Gaussian Smoothing)을 적용하여 데이터의 희소성 문제를 줄였다. 실험 결과에서 본 방법이 기존의 방법보다 성능이 좋을 뿐만 아니라 운율 경계를 추정하기 위한 자질을 독립시켰기 때문에 다른 시스템과의 호환성도 높다는 것을 알 수 있었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10d
• /
• pp.535-537
• /
• 2002

본 논문은 실시간 카메라 입력 환경에서의 새로운 얼굴 검출 및 추적 알고리즘을 제안한다. 복잡한 배경과 다양한 조명 조건에 관계 없이 얼굴을 검출하고 추적하기 위해 세 종류의 웨이블릿 변환된 형판을 사용하고 특히 다양한 조명 조건을 극복하기 위해 최소-최대 정규화(Min-Max Normalization)와 히스토그램 평활화를 혼합 적용하여 매우 밝거나, 매우 어두운 영상에서의 얼굴 오 검출 및 놓침을 줄이도록 하였다. 또한 세가지 크기의 얼굴 형판을 이용함으로써 입력 영상에 존재하는 다양한 크기의 얼굴도 검출할 수 있었으며, 효과적인 얼굴 추적 기법을 통해 다음 프레임에서의 얼굴 위치를 예측하여 그 지점에서의 탐색 영역에 형판 정합을 수행함으로써 수행 시간도 단축시킬 수 있었다. 실험을 위해 다양한 조명 조건에 따라 여섯 종류로 분류한 동영상 데이터에서 제안한 알고리즘은 약 96.8%의 뛰어난 얼굴 검출율을 보여 주었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04a
• /
• pp.271-273
• /
• 2000

본 논문에서는 TCP 수신자가 지연 ACK 옵션(Delayed ACK Option)을 사용할 경우에 TCP 송신자에게 발생하는 성능 저하요인들을 분석하고 다음과 같은 해결책을 제시하였다. 먼저, 느린 시작 구간(Slow Start phase) 처음에 생기는 ACK 타임아웃은 큰 초기 윈도우(large initial window)또는 1-bit 마킹 기법을 통해 해결할 수 있다. 그리고, 느린 시작 구간과 혼잡 회피 구간(Congestion Avoidance phase)에서 혼잡 윈도우(cwnd)가 천천히 증가하는 문제는 적절히 바이트 카운팅 기법을 사용함으로써 해결할 수 있다. 마지막으로, 송신자가 버스트(burst)한 데이터를 네트웍에 발생시키는 문제는 트래픽을 평활(pacing)함으로써 해결할 수 있다. 또한 본 연구에서는 분석적 모델링을 통하여 TCP가 보내는 평균 전송률을 구하였으며 이 결과는 TCP에 친화한 전송률 기반 전송방법(TCP Friendly Rate Based Control)에 응용될 수 있을 것이다. 그리고 시뮬레이션을 통해서 제시한 방법의 성능이 향상됨을 확인하였다.