• 한국수자원학회논문집
• /
• 제54권9호
• /
• pp.643-656
• /
• 2021

최근 국내 상수도 관리 기술은 고도로 발달하고 있으며, 이 과정에서 상수관망 내 용수공급 현황을 파악하고 예측하기 위한 컴퓨터 수리·수질 해석 모형은 핵심적인 역할을 수행하고 있다. 그러나 대규모 네트워크의 경우 컴퓨터 해석모형의 부담을 가중하고, 특히 짧은 계산시간 간격과 긴 모의 시간이 요구되는 수질해석의 경우, 막대한 계산시간이 소요되어 다양한 수질모의 및 분석이 어려운 경우가 발생한다. 본 연구에서는 대규모 상수관망시스템의 수질해석의 계산효율을 개선하기 위해 상수도 공급계통을 2단계로 계층화한 후, 계층화된 네트워크를 대상으로 수질모의를 수행하는 계층적 수질모의 기법을 제안하였다. 제안된 모의기법은 국내 대규모 상수도 네트워크에 적용하였으며, 다양한 염소투입농도 시나리오에 따른 잔류염소농도의 시공간적 분포를 모의하고 분석한 결과를 제시하였다.

• 전자공학회논문지
• /
• 제54권5호
• /
• pp.35-41
• /
• 2017

본 논문에서는 지식추출(knowledge distillation) 및 지식전달(knowledge transfer)을 위하여 최근에 소개된 선생-학생 프레임워크 기반의 힌트(Hint)-knowledge distillation(KD) 학습기법에 대한 성능을 분석한다. 본 논문에서 고려하는 선생-학생 프레임워크는 현재 최신 딥러닝 모델로 각광받고 있는 딥 residual 네트워크를 이용한다. 따라서, 전 세계적으로 널리 사용되고 있는 오픈 딥러닝 프레임워크인 Caffe를 이용하여 학생모델의 인식 정확도 관점에서 힌트-KD 학습 시 선생모델의 완화상수기반의 KD 정보 비중에 대한 영향을 살펴본다. 본 논문의 연구결과에 따르면 KD 정보 비중을 단조감소하는 경우보다 초기에 설정된 고정된 값으로 유지하는 것이 학생모델의 인식 정확도가 더 향상된다는 것을 알 수 있었다.

• 대한전자공학회논문지TC
• /
• 제46권8호
• /
• pp.43-49
• /
• 2009

본 논문에서는 에러율이 높은 무선 센서 네트워크의 수명을 연장시키기 위한 라우팅 알고리즘을 제안한다 제안하는 기법은 무선 링크의 품질을 기반으로 하여 패킷을 송수신한 이후의 잔여 에너지를 미리 추정하고, 그 때의 에너지가 가장 많이 남아 있을 것으로 예상되는 경로로 데이터를 전송한다. 제안하는 기법은 에러율을 고려하기 때문에 불필요한 재전송에 따른 에너지 소비를 줄이고 트래픽도 골고루 분산시킨다. 그리고 송수신 이후의 잔여 에너지양들 중에서 최소값이 가장 클 것으로 예상되는 경로를 선택함으로써 노드의 에너지 고갈을 최대한으로 지연시킨다. 다른 방식들에 비하여 제안하는 기법이 네트워크의 수명을 더욱 연장시킨다는 사실을 실험을 통하여 확인할 수 있었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권1호
• /
• pp.68-86
• /
• 2015

Future wireless network aims to integrate different radio access networks (RANs) to provide a seamless access and service continuity. In this paper, a new resource denotation method is proposed in the WLAN and LTE heterogeneous networks based on a concept of spectral bandwidth mapping. This method simplifies the denotation of system resources and makes it possible to calculate system residual capacity, upon which an economic model-based network selection algorithm is designed in both under-loaded and over-loaded scenarios in the heterogeneous networks. The simulation results show that this algorithm achieves better performance than the utility function-based access selection (UFAS) method proposed in [12] in increasing system capacity and system revenue, achieving load balancing and reducing the new call blocking probability in the heterogeneous networks.

• Journal of Information Processing Systems
• /
• 제7권1호
• /
• pp.29-42
• /
• 2011

Wireless sensor networks are composed of a large number of sensor nodes with limited energy resources. One critical issue in wireless sensor networks is how to gather sensed information in an energy efficient way, since their energy is limited. The clustering algorithm is a technique used to reduce energy consumption. It can improve the scalability and lifetime of wireless sensor networks. In this paper, we introduce a clustering protocol with mode selection (CPMS) for wireless sensor networks. Our scheme improves the performance of BCDCP (Base Station Controlled Dynamic Clustering Protocol) and BIDRP (Base Station Initiated Dynamic Routing Protocol) routing protocol. In CPMS, the base station constructs clusters and makes the head node with the highest residual energy send data to the base station. Furthermore, we can save the energy of head nodes by using the modes selection method. The simulation results show that CPMS achieves longer lifetime and more data message transmissions than current important clustering protocols in wireless sensor networks.

• 한국시뮬레이션학회:학술대회논문집
• /
• 한국시뮬레이션학회 2000년도 추계학술대회 논문집
• /
• pp.97-101
• /
• 2000

Energy consumption is considered as a principal ingredient in mobile wireless ad-hoc networks. In such a network, most of mobile nodes takes a role in forwarding messages received from neighbor nodes. Energy of these nodes is consumed in different rates depending on message traffic routes. This paper proposes a scheme to balance routing energy consumption by transferring routing function from node with small residual energy to node with enough residual energy. This scheme requires additional local message transfer, increasing the energy consumption of nodes to transfer routing function, and increasing total energy consumption of ad-hoc network. But balancing of energy consumption make the system lifetime the longer and increase the average node lifetime.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2019년도 춘계학술발표대회
• /
• pp.525-528
• /
• 2019

Atmospheric particle removal is a challenging task and attacks wide interests in computer vision filed. In this paper, we proposed a single image snow removal framework based on deep residual networks. According to the fact that there are various snow sizes in a snow image, the inception module which consists of different filter kernels was adopted to extract multiple resolution features of the input snow image. Except the traditional mean square error loss, the perceptual loss and total variation loss were employed to generate more clean images. Experimental results on synthetic and realistic snow images indicated that the proposed method achieves superior performance in respect of visual perception and objective evaluation.

• ETRI Journal
• /
• 제41권3호
• /
• pp.326-336
• /
• 2019

A wireless power transfer technique can solve the power capacity problem in wireless rechargeable sensor networks (WRSNs). The charging strategy is a wide-spread research problem. In this paper, we propose a demand-based charging strategy (DBCS) for WRSNs. We improved the charging programming in four ways: clustering method, selecting to-be-charged nodes, charging path, and charging schedule. First, we proposed a multipoint improved K-means (MIKmeans) clustering algorithm to balance the energy consumption, which can group nodes based on location, residual energy, and historical contribution. Second, the dynamic selection algorithm for charging nodes (DSACN) was proposed to select on-demand charging nodes. Third, we designed simulated annealing based on performance and efficiency (SABPE) to optimize the charging path for a mobile charging vehicle (MCV) and reduce the charging time. Last, we proposed the DBCS to enhance the efficiency of the MCV. Simulations reveal that the strategy can achieve better performance in terms of reducing the charging path, thus increasing communication effectiveness and residual energy utility.

• 전기학회논문지
• /
• 제68권1호
• /
• pp.98-101
• /
• 2019

In this paper, we propose the power-based pipelined-forwarding MAC protocol which can select relay nodes according to the residual power and energy harvesting rate in EH-WSN (energy-harvesting wireless sensor networks). The proposed MAC follows a pipelined-forwarding scheme in which nodes repeatedly sleep and wake up in an EH-WSN environment and data is continuously transmitted from a high-level node to a low-level node. The sleep interval is adaptively controlled so that nodes with low energy harvesting rate can be charged sufficiently, thereby minimizing the transmission delay and increasing the network lifetime. Simulation shows that the proposed MAC protocol improves the balance of residual power and network lifetime.

• Journal of Information Processing Systems
• /
• 제17권2호
• /
• pp.385-398
• /
• 2021

The complexity of deep learning models affects the real-time performance of gesture recognition, thereby limiting the application of gesture recognition algorithms in actual scenarios. Hence, a residual learning neural network based on a deep convolutional neural network is proposed. First, small convolution kernels are used to extract the local details of gesture images. Subsequently, a shallow residual structure is built to share weights, thereby avoiding gradient disappearance or gradient explosion as the network layer deepens; consequently, the difficulty of model optimisation is simplified. Additional convolutional neural networks are used to accelerate the refinement of deep abstract features based on the spatial importance of the gesture feature distribution. Finally, a fully connected cascade softmax classifier is used to complete the gesture recognition. Compared with the dense connection multiplexing feature information network, the proposed algorithm is optimised in feature multiplexing to avoid performance fluctuations caused by feature redundancy. Experimental results from the ISOGD gesture dataset and Gesture dataset prove that the proposed algorithm affords a fast convergence speed and high accuracy.