• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.10a
• /
• pp.27-30
• /
• 2001

기존의 오프라인에서 이루어지는 데이터 분석 방법은 대부분 수작업으로 이루어지기 때문에 데이터를 분석하는데 시간과 장소의 제약을 받으며 데이터 수집 과정에서 오류 발생률이 높게 나타나고, 예전보다 분석해야 할 데이터의 양이 방대해짐에 따라 사용자는 더 많은 시간과 노력을 할애해야 한다. 이에 데이터를 신속, 정확하게 분석하기 위해서는 보다 향상된 데이터 분석 기법의 개발이 필요하다. 따라서 본 논문에서는 이러한 문제점을 해결함과 동시에 온라인 환경에서 사용자가 데이터를 효율적으로 입력하고 분석할 수 있는 다중항목 데이터 분석 시스템을 제안한다. 이 시스템으로 사용자는 현재 일반화되어 있는 웹을 통하여 분석된 데이터 결과를 제공받을 수 있다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.2
• /
• pp.63-72
• /
• 2011

Broadcast-based data dissemination has become a widely accepted approach of communication in the mobile computing environment. However, with a large set of data items, the expected delay of receiving a desired data increases due to the sequential nature of the broadcast channel. With the objective of minimizing this wait time, this paper explores the problem of data broadcast over multiple channels. In traditional approaches, data items are partitioned based on their access probabilities and allocated on multiple channels, assuming flat data scheduling per channel. If the data items allocated on the same channel are broadcast in different frequencies based on their access probabilities, the performance will be enhanced further. In this respect, this paper proposes a new broadcast scheduling scheme named two level dynamic programming(TLDP) which can reflect a variation of access probabilities among data items allocated on the same channel.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.17 no.7
• /
• pp.1597-1601
• /
• 2013

Multiple antenna technique is attracting attention as a core technology for next-generation mobile communication systems to accommodate explosively increasing mobile data traffic. Especially, recent researches focus on multi-user multiple input multiple output (MU-MIMO) system where base stations are equipped with several tens of transmit antennas and transmit data to multiple terminals (users) simultaneously. To enhance the performance of MU-MIMO systems, we, in this paper, propose an adaptive user selection algorithm which adaptively selects a user set according to varying channel states. According to Monte-Carlo based computer simulations, the performance of proposed scheme is significantly improved compared to the conventional scheme without user selection and approaches that of exhaustive search-based optimal scheme. On the other hand, the proposed scheme can reduce the computational complexity to $K/(2^K-1)$ compared to the optimal scheme where K denotes the number of total users.

• Journal of KIISE
• /
• v.43 no.6
• /
• pp.686-691
• /
• 2016

In this paper, we introduced the adaptive spectral co-clustering, a spectral clustering for multiview data, especially data with more than three views. In the adaptive spectral co-clustering, the performance is improved by sharing information from diverse views. For the efficiency in information sharing, a co-training approach is adopted. In the co-training step, a set of parameters are estimated to make all views in data maximally independent, and then, information is shared with respect to estimated parameters. This co-training step increases the efficiency of information sharing comparing with ordinary feature concatenation and co-training methods that assume the independence among views. The adaptive spectral co-clustering was evaluated with synthetic dataset and multi lingual document dataset. The experimental results indicated the efficiency of the adaptive spectral co-clustering with the performances in every iterations and similarity matrix generated with information sharing.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.4
• /
• pp.111-117
• /
• 2011

In this paper, we propose an QoS aware multi-layer MAC(QAML-MAC) protocol in a wireless sensor networks. Since the proposed protocol is based on the sleep-awake architecture, which save node's energy to prolong the entire network lifetime. For this purpose the QAML-MAC first classifies incoming data according to their transmission urgency and then saves them. The protocol also adapts the cross-layer concept to re-arrange the order of transmission with the same destination. So the delay can be decreased, which can not be obtained with the previous related protocols. And high priority data such as real-time multimedia or critical value in the field monitoring applications can be transmitted quickly, Furthermore the proposed protocol has advantage of decreasing transmitted data collisions using multiple layers of idle listening when there is no high-priority data. So energy consumptions of sensor nodes can be saved and the network lifetime can be prolonged.

• Journal of Digital Convergence
• /
• v.13 no.1
• /
• pp.263-268
• /
• 2015

Put your information in the object-based sensors and mobile networks has been developed that correlate with ubiquitous information technology as the development of IT technology. However, a security solution is to have the data stored in the server, what minimal conditions. In this paper, we propose a data management method is applied to a hash chain of the properties of the multiple techniques to the data used by the big user and the data services to ensure safe handling large amounts of data being provided in the big data services. Improves the safety of the data tied to the hash chain for the classification to classify the attributes of the data attribute information according to the type of data used for the big data services, functions and characteristics of the proposed method. Also, the distributed processing of big data by utilizing the access control information of the hash chain to connect the data attribute information to a geographically dispersed data easily accessible techniques are proposed.

• Proceedings of the Korean Information Science Society Conference
• /
• 2002.10d
• /
• pp.253-255
• /
• 2002

본 논문에서는 다중 모델 기계학습 기법을 이용하여 문서 자동 분류의 성능과 신뢰도를 향상시킬 수 있는 연구와 실험 결과를 기술하였다. 기존의 다중 모텔 기계 학습법들이 훈련 데이터 또는 학습 알고리즘의 편향에 의한 오류를 극복하고 한 것들인데 비해 본 논문에서 제안한 메타 학습을 이용한 하이브리드 다중 모델 방식은 이 두 가지의 오류 원인을 동시에 해소하고자 하였다. 다양한 문서 집합에 대한 실험 결과, 본 연구에서 제안한 하이브리드 다중 모델 학습법이 전반적으로 기존의 일반 다중모델 학습법들에 비해 높은 성능을 보였으며, 다중 모델의 결합 방식으로서 메타 학습이 투표 방식에 비해 효율적인 것으로 나타났다.

• Journal of Digital Convergence
• /
• v.13 no.10
• /
• pp.271-278
• /
• 2015

Recently, bit data is difficult to easily collect the desired data because big data is collected via the Internet. Big data is higher than the rate at which the data type and the period of time for which data is collected depending on the size of data increases. In particular, since the data of all different by the intended use and the type of data processing accuracy and computational cost is one of the important items. In this paper, we propose data processing method using a dual-chain in a manner to minimize the computational cost of the data when data is correctly extracted at the same time a multi-layered process through the desired number of the user and different kinds of data on the Internet. The proposed scheme is classified into a hierarchical data in accordance with the intended use and method to extract various kinds of data. At this time, multi-processing and tie the data hash with the double chain to enhance the accuracy of the reading. In addition, the proposed method is to organize the data in the hash chain for easy access to the hierarchically classified data and reduced the cost of processing the data. Experimental results, the proposed method is the accuracy of the data on average 7.8% higher than conventional techniques, processing costs were reduced by 4.9% of the data.

• Journal of Navigation and Port Research
• /
• v.36 no.8
• /
• pp.631-635
• /
• 2012

Underwater acoustic communication has multipath error because of reflection by sea-level and sea-bottom. The multipath of underwater channel causes signal distortion and error floor. In this paper, in order to design an efficient packet structure, we employ channel coding scheme and phase recovery algorithm. For channel coding scheme, half rate LDPC channel coding scheme with N=1944 and K=972 was used. Also, decision directed phase recovery was used for correcting phase offset induced by multipath. Based on these algorithms, we propose length of data for optimal packet structure in the environment of oceanic experimentation.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.215-220
• /
• 2021

다중 도메인에 대해 답변 생성 모델이 동작 가능하도록 하는 가장 쉬운 방법은 모든 도메인의 데이터를 순서와 상관없이 한번에 학습하는 것이다. 하지만 이경우, 발화에 상관 없이 지나치게 일반적인 답변을 생성하는 문제가 발생한다. 이에 반해, 도메인을 분리하여 도메인을 순차적으로 학습할 경우 일반적인 답변 생성 문제를 해결할 수 있다. 하지만 이경우 새로운 도메인의 데이터를 학습할 때, 기존에 학습한 도메인에 대한 성능이 저하되는 파괴적 망각 현상이 발생한다. 파괴적 망각 현상을 해결하기 위하여 다양한 지속학습기법이 제안되었으며, 그 중 메모리 리플레이 방법은 새로운 도메인 학습시 기존 도메인의 데이터를 함께 학습하는 방법으로 파괴적 망각 현상을 해결하고자 하였다. 본 논문에서는, 사람의 기억 시스템에 대한 모형인 앳킨슨-쉬프린 기억 모형에서 착안하여 사람이 기억을 저장하는것과 유사한 방법으로 메모리 리플레이 방법의 메모리 관리방법을 제안하였고, 해당 메모리 관리법을 활용하는 메모리 리플레이 방법을 통해 답변 생성 모델의 파괴적 망각 현상을 줄이고자 하였다. 다중 도메인 답변 생성에 대한 데이터셋인 MultiWoZ-2.0를 사용하여 제안 모델을 학습 및 평가하였고, 제안 모델이 다중 도메인 답변 생성 모델의 파괴적 망각 현상을 감소시킴을 확인하였다.