• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.27 no.3
• /
• pp.106-112
• /
• 2004

This paper presents a new heuristic algorithm for the machine cell(MC) formation problem. MC formation problem is represented as an unbalanced k-way network partition and the proposed algorithm uses four stage-approach to solve the problem. Four stages are natural sub-network formation, determination of intial vertexes for each sub-network, determination of initial partition, and improvement of initial partition. Results of experiments show that the suggested algorithm provides near optimal solutions within very short computational time.

• The Journal of Industrial Distribution & Business
• /
• v.9 no.5
• /
• pp.25-35
• /
• 2018

Purpose - The purpose of this paper is to partition a last-mile delivery network into zones and to determine locations of last mile delivery centers (LMDCs) in Bangkok, Thailand. Research design, data, and methodology - As online shopping has become popular, parcel companies need to improve their delivery services as fast as possible. A network partition has been applied to evaluate suitable service areas by using METIS algorithm to solve this scenario and a facility location problem is used to address LMDC in a partitioned area. Research design, data, and methodology - Clustering and mixed integer programming algorithms are applied to partition the network and to locate facilities in the network. Results - Network partition improves last mile delivery service. METIS algorithm divided the area into 25 partitions by minimizing the inter-network links. To serve short-haul deliveries, this paper located 96 LMDCs in compact partitioning to satisfy customer demands. Conclusions -The computational results from the case study showed that the proposed two-phase algorithm with network partitioning and facility location can efficiently design a last-mile delivery network. It improves parcel delivery services when sending parcels to customers and reduces the overall delivery time. It is expected that the proposed two-phase approach can help parcel delivery companies minimize investment while providing faster delivery services.

• Proceedings of the KIEE Conference
• /
• 2000.11a
• /
• pp.223-225
• /
• 2000

Network congestion is important problem that must be managed for active power contracts in deregulated power industry. Existing network partition technique is based on statistical method don't suggest clear network partition criterion. So in this paper we proposed a new network partition technique using the marginal cost sensitivity to fairly allocate congestion cost to network user.

• Journal of Korea Spatial Information System Society
• /
• v.9 no.3
• /
• pp.35-50
• /
• 2007

In a data centric sensor network, a sensor node to store data is determined by the measured data value of each sensor node. Therefore, if the same data occur frequently, the energy of the sensor node to store the data is exhausted quickly due to the concentration of loads. And if the sensor network is extended, the communication cost for storing data and processing queries is increased, since the length of the routing path for them is usually in the distance. However, the existing researches that generally focus on the efficient management of data storing can not solve these problems efficiently. In this paper, we propose a NUNS(Non-Uniform Network Split) method that can distribute loads of sensor nodes and decrease the communication cost caused by the sensor network extension. By dividing the sensor network into non-uniform partitions that have the minimum difference in the number of sensor nodes and the splitted area size and storing the data which is occurred in a partition at the sensor nodes within the partition, the NUNS can distribute loads of sensor nodes and decrease the communication cost efficiently. In addition, by dividing each partition into non-uniform zones that have the minimum difference in the splitted area size as many as the number of the sensor nodes in the partition and allocating each of them as the processing area of each sensor node, the NUNS can protect a specific sensor node from the load concentration and decrease the unnecessary routing cost.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.11
• /
• pp.4483-4501
• /
• 2020

Recent findings identified the scatter and shape of MFD (macroscopic fundamental diagram) is heavily influenced by the spatial distribution of link density in a road network. This implies that the concept of MFD can be utilized to divide a heterogeneous road network with different degrees of congestion into multiple homogeneous subnetworks. Considering the actual traffic data is usually incomplete and inaccurate while most traffic partition algorithms rely on the completeness of the data, we proposed a three-step partitioned algorithm called Iso-MB (Isoperimetric algorithm - Merging - Boundary adjustment) permitting of incompletely input data in this paper. The proposed algorithm was implemented and verified in a simulated urban transportation network. The existence of well-defined MFD in each subnetwork was revealed and discussed and the selection of stop parameter in the isoperimetric algorithm was explained and dissected. The effectiveness of the approach to the missing input data was also demonstrated and elaborated.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.28 no.4
• /
• pp.131-143
• /
• 2003

In this paper, we present a survey about the various graph partition problems including the clustering problem, the k-cut problem, the multiterminal cut problem, the multicut problem, the sparsest cut problem, the network attack problem, the network disconnection problem. We compare those problems focusing on the problem characteristics such as the objective function and the conditions that the partitioned clusters should satisfy. We also introduce the mathematical programming formulations, and the solution approaches developed for the problems.

• 한국공간정보시스템학회:학술대회논문집
• /
• 2007.06a
• /
• pp.59-64
• /
• 2007

데이타 중심 센서 네트워크에서는 측정된 데이타의 값에 따라 데이타를 저장하는 센서 노드가 결정되기 때문에 같은 값을 갖는 데이타가 빈번하게 발생하면 이를 저장하는 센서 노드에 부하가 집중되어 에너지가 빠르게 고갈되는 문제가 있다. 또한 센서 네트워크가 확장되면 데이타 저장 및 질의 처리시 목적 센서 노드로의 라우팅 거리가 멀어져 통신 비용이 증가되는 문제가 있다. 그러나 기존 연구들은 데이타 저장의 효율적인 관리에만 치우쳐 이와 같은 문제를 효율적으로 해결하지 못하고 있다. 본 논문에서는 이러한 문제를 해결하기 위해 비균등 네트워크 분할(Non-Uniform Network Spilt: NUNS) 기법을 제안한다. NUNS는 센서 네트워크를 센서 노드 개수와 분할된 영역 크기 차이가 최소가 되도록 비균등 크기의 Partition으로 분할하고 각 Partition에서 발생한 데이타를 그 Partition 내의 센서 노드가 저장 관리함으로써 센서 노드의 데이타 저장 부하를 분산시키고, 센서 네트워크의 확장에 따른 통신 비용을 줄인다. 그리고 NUNS는 각 Partition을 분할된 영역 크기 차이가 최소가 되도록 센서 노드 개수만큼 비균등하게 Zone으로 분할함으로써 센서 노드가 없는 Zone으로 인해 센서 노드에 부하가 집중되는 것을 막고 불필요한 라우팅 비용을 줄인다.

• ETRI Journal
• /
• v.45 no.2
• /
• pp.318-328
• /
• 2023

Recently, embedded systems, such as mobile platforms, have multiple processing units that can operate in parallel, such as centralized processing units (CPUs) and neural processing units (NPUs). We can use deep-learning compilers to generate machine code optimized for these embedded systems from a deep neural network (DNN). However, the deep-learning compilers proposed so far generate codes that sequentially execute DNN operators on a single processing unit or parallel codes for graphic processing units (GPUs). In this study, we propose PartitionTuner, an operator scheduler for deep-learning compilers that supports multiple heterogeneous PUs including CPUs and NPUs. PartitionTuner can generate an operator-scheduling plan that uses all available PUs simultaneously to minimize overall DNN inference time. Operator scheduling is based on the analysis of DNN architecture and the performance profiles of individual and group operators measured on heterogeneous processing units. By the experiments for seven DNNs, PartitionTuner generates scheduling plans that perform 5.03% better than a static type-based operator-scheduling technique for SqueezeNet. In addition, PartitionTuner outperforms recent profiling-based operator-scheduling techniques for ResNet50, ResNet18, and SqueezeNet by 7.18%, 5.36%, and 2.73%, respectively.

• Journal of Institute of Control, Robotics and Systems
• /
• v.8 no.2
• /
• pp.120-125
• /
• 2002

An optimization approach is used to partition the field of view. A cost function is defined to represent the constraints on the solution, which is then mapped onto a two-dimensional Hopfield neural network for minimization. Each neuron in the network represents a possible match between a field of view and one or multiple objects. Partition is achieved by initializing each neuron that represents a possible match and then allowing the network to settle down into a stable state. The network uses the initial inputs and the compatibility measures between a field of view and one or multiple objects to find a stable state.

• Proceedings of the KSME Conference
• /
• 2001.11a
• /
• pp.667-672
• /
• 2001

An optimization approach is used to partition the field of view. A cost function is defined to represent the constraints on the solution, which is then mapped onto a two-dimensional Hopfield neural network for minimization. Each neuron in the network represents a possible match between a field of view and one or multiple objects. Partition is achieved by initializing each neuron that represents a possible match and then allowing the network to settle down into a stable state. The network uses the initial inputs and the compatibility measures between a field of view and one or multiple objects to find a stable state.