Earticle

본 논문에서는 퍼지 클러스터를 이용한 비선형 추론을 위한 퍼지 추론 시스템을 소개한다. 전형적으로, 비선형 추론을 위한 퍼지 규칙의 생성은 일반적으로 입력 벡터 차원이 증가하면 규칙의 수가 지수적으로 증가하게 된다. 이러한 문제점을 해결하기 위해, 퍼지 클러스터를 표현할 수 있는 퍼지 클러스터링 알고리즘을 이용하여 입력 벡터 공간을 분산 형태로 분할하여 퍼지 모델의 규칙을 설계한다. 이러한 방법으로 복잡하고 비선형적인 공정을 퍼지 모델링 할 수 있다. 퍼지 규칙의 전반부는 퍼지 클러스터를 갖는 FCM 클러스터링 알고리즘에 의해 결정된다. 퍼지 규칙의 후반부는 4가지 형태의 다항식 함수의 형태를 가지며, 각 규칙의 후반부 파라미터들은 표준 최소자승법을 이용함으로써 추정된다. 그리고 비선형 공정의 특성 및 성능을 평가하기 위하여 비선형 공정으로 많이 이용되고 있는 데이터를 이용한다. 실험 결과는 비선형 추론이 가능하다는 것을 보여준다.

In this paper, we introduce a fuzzy inference systems for nonlinear inference using fuzzy cluster. Typically, the generation of fuzzy rules for nonlinear inference causes the problem that the number of fuzzy rules increases exponentially if the input vectors increase. To handle this problem, the fuzzy rules of fuzzy model are designed by dividing the input vector space in the scatter form using fuzzy clustering algorithm which expresses fuzzy cluster. From this method, complex nonlinear process can be modeled. The premise part of the fuzzy rules is determined by means of FCM clustering algorithm with fuzzy clusters. The consequence part of the fuzzy rules have four kinds of polynomial functions and the coefficient parameters of each rule are estimated by using the standard least-squares method. And we use the data widely used in nonlinear process for the performance and the nonlinear characteristics of the nonlinear process. Experimental results show that the non-linear inference is possible.

In clustering-based approaches, cluster heads closer to the sink are usually burdened with much more relay traffic and thus, tend to die early. To address this problem, distance-aware clustering approaches, such as energy-efficient unequal clustering (EEUC), that adjust the cluster size according to the distance between the sink and each cluster head have been proposed. However, the network lifetime of such approaches is highly dependent on the distribution of the sensor nodes, because, in randomly distributed sensor networks, the approaches do not guarantee that the cluster energy consumption will be proportional to the cluster size. To address this problem, we propose a novel approach called CACD (Clustering Algorithm Considering node Distribution), which is not only distance-aware but also node density-aware approach. In CACD, clusters are allowed to have limited member nodes, which are determined by the distance between the sink and the cluster head. Simulation results show that CACD is 20%-50% more energy-efficient than previous work under various operational conditions considering the network lifetime.

Although density peak clustering can often easily yield excellent results, there is still room for improvement when dealing with complex, high-dimensional datasets. One of the main limitations of this algorithm is its reliance on geometric distance as the sole similarity measurement. To address this limitation, we draw inspiration from the information bottleneck theory, and propose a novel density peak clustering algorithm that incorporates this theory as a similarity measure. Specifically, our algorithm utilizes the joint probability distribution between data objects and feature information, and employs the loss of mutual information as the measurement standard. This approach not only eliminates the potential for subjective error in selecting similarity method, but also enhances performance on datasets with multiple centers and high dimensionality. To evaluate the effectiveness of our algorithm, we conducted experiments using ten carefully selected datasets and compared the results with three other algorithms. The experimental results demonstrate that our information bottleneck-based density peaks clustering (IBDPC) algorithm consistently achieves high levels of accuracy, highlighting its potential as a valuable tool for data clustering tasks.

Fuzzy co-clustering is sensitive to noise data. To overcome this noise sensitivity defect, possibilistic clustering relaxes the constraints in FCM-type fuzzy (co-)clustering. In this paper, we introduce a new possibilistic fuzzy co-clustering algorithm based on information bottleneck (ibPFCC). This algorithm combines fuzzy co-clustering and possibilistic clustering, and formulates an objective function which includes a distance function that employs information bottleneck theory to measure the distance between feature data point and feature cluster centroid. Many experiments were conducted on three datasets and one artificial dataset. Experimental results show that ibPFCC is better than such prominent fuzzy (co-)clustering algorithms as FCM, FCCM, RFCC and FCCI, in terms of accuracy and robustness.

In this paper, an improved automated spectral clustering (IASC) algorithm is proposed to address the limitations of the traditional spectral clustering (TSC) algorithm, particularly its inability to automatically determine the number of clusters. Firstly, a cluster number evaluation factor based on the optimal clustering principle is proposed. By iterating through different k values, the value corresponding to the largest evaluation factor was selected as the first-rank number of clusters. Secondly, the IASC algorithm adopts a density-sensitive distance to measure the similarity between the sample points. This rendered a high similarity to the data distributed in the same high-density area. Thirdly, to improve clustering accuracy, the IASC algorithm uses the cosine angle classification method instead of K-means to classify the eigenvectors. Six algorithms-K-means, fuzzy C-means, TSC, EIGENGAP, DBSCAN, and density peak-were compared with the proposed algorithm on six datasets. The results show that the IASC algorithm not only automatically determines the number of clusters but also obtains better clustering accuracy on both synthetic and UCI datasets.

Recently, as data demand has increased owing to the rapidly increasing demand for wireless devices and the influence of data traffic, various technologies are being developed to support it. Among them, millimeter-wave (mmWave) frequencies with rich spectra and high data-transmission rates suffer from the problem of large path loss. Accordingly, there is a growing interest in unmanned aerial vehicles (UAVs) and reconfigurable intelligent surfaces (RISs), which can be utilized advantageously to reconstruct wireless communication environments. Therefore, this work considers a large-scale system comprising a number of users and Flying RISs, combining UAVs and RISs to increase algorithm utilization. We propose a deep neural network-based algorithm that places Flying RISs in an appropriate location so that they can support as many users as possible. Simulation results confirmed that the proposed technique could place Flying RISs in an efficient location with higher accuracy and speed in large-scale systems compared to existing techniques.

Responding to changes in artificial intelligence models and the data environment is crucial for increasing data-learning accuracy and inference stability of industrial applications. A learning model that is overfitted to specific training data leads to poor learning performance and a deterioration in flexibility. Therefore, an early stopping technique is used to stop learning at an appropriate time. However, this technique does not consider the homogeneity and independence of the data collected by heterogeneous nodes in a differential network environment, thus resulting in low learning accuracy and degradation of system performance. In this study, the generalization performance of neural networks is maximized, whereas the effect of the homogeneity of datasets is minimized by achieving an accuracy of 99.7%. This corresponds to a decrease in delay time by a factor of 2.33 and improvement in performance by a factor of 2.5 compared with the conventional method.

The interaction of power load information provides reliable data support for accessing user-side electrical energy storage devices and distributed renewable energy sources. However, owing to the large volume of interactive information and the numerous security threats faced during the interaction, anomaly detection has become one of the most challenging problems in smart grids. To address this issue, an anomaly detection method was developed that consists of three stages. First, feature extraction is performed based on the power load information. Then, a robust principal component analysis method is used for the preliminary classification of the extracted features. Finally, an improved K-means clustering algorithm is employed to refine the classification results into completely non-overlapping groups and detect anomalies from the classified data. Experimental results demonstrate that the proposed method can effectively and accurately detect anomalies from power load data.

In this paper, we present a virtual laboratory platform (VLP) baptized Mercury allowing students to make practical work (PW) on different aspects of mobile wireless sensor networks (WSNs). Our choice of WSNs is motivated mainly by the use of real experiments needed in most courses about WSNs. These experiments require an expensive investment and a lot of nodes in the classroom. To illustrate our study, we propose a course related to energy efficient and safe weighted clustering algorithm. This algorithm which is coupled with suitable routing protocols, aims to maintain stable clustering structure, to prevent most routing attacks on sensor networks, to guaranty energy saving in order to extend the lifespan of the network. It also offers a better performance in terms of the number of re-affiliations. The platform presented here aims at showing the feasibility, the flexibility and the reduced cost of such a realization. We demonstrate the performance of the proposed algorithms that contribute to the familiarization of the learners in the field of WSNs.

This research proposes a new strategy where documents are encoded into string vectors and modified version of k means algorithm to be adaptable to string vectors for text clustering. Traditionally, when k means algorithm is used for pattern classification, raw data should be encoded into numerical vectors. This encoding may be difficult, depending on a given application area of pattern classification. For example, in text clustering, encoding full texts given as raw data into numerical vectors leads to two main problems: huge dimensionality and sparse distribution. In this research, we encode full texts into string vectors, and modify the k means algorithm adaptable to string vectors for text clustering.

Centralized hierarchical routing protocols are often used to solve the problems of uneven energy consumption and short network life in wireless sensor networks (WSNs). Clustering and cluster head election have become the focuses of WSNs. In this paper, an energy balanced clustering routing algorithm optimized by sine cosine algorithm (SCA) is proposed. Firstly, optimal cluster head number per round is determined according to surviving node, and the candidate cluster head set is formed by selecting high-energy node. Secondly, a random population with a certain scale is constructed to represent a group of cluster head selection scheme, and fitness function is designed according to inter-cluster distance. Thirdly, the SCA algorithm is improved by using monotone decreasing convex function, and then a certain number of iterations are carried out to select a group of individuals with the minimum fitness function value. From simulation experiments, the process from the first death node to 80% only needs about 30 rounds. This improved algorithm balances the energy consumption among nodes and avoids premature death of some nodes. And it greatly improves the energy utilization and extends the effective life of the whole network.

In the conventional clustering algorithms, an object could be assigned to only one group. However, this is sometimes not the case in reality, there are cases where the data do not belong to one group. As against, the fuzzy clustering takes into consideration the degree of fuzzy membership of each pixel relative to different classes. In order to overcome some shortcoming with traditional clustering methods, such as slow convergence and their sensitivity to initialization values, we have used the Harmony Search algorithm. It is based on the population metaheuristic algorithm, imitating the musical improvisation process. The major thrust of this algorithm lies in its ability to integrate the key components of population-based methods and local search-based methods in a simple optimization model. We propose in this paper a new unsupervised clustering method called the Fuzzy Harmony Search-Fourier Transform (FHS-FT). It is based on hybridization fuzzy clustering and the harmony search algorithm to increase its exploitation process and to further improve the generated solution, while the Fourier transform to increase the size of the image's data. The results show that the proposed method is able to provide viable solutions as compared to previous work.

Clustering is a NP-hard problem that is used to find the relationship between patterns in a given set of patterns. It is an unsupervised technique that is applied to obtain the optimal cluster centers, especially in partitioned based clustering algorithms. On the other hand, cat swarm optimization (CSO) is a new meta-heuristic algorithm that has been applied to solve various optimization problems and it provides better results in comparison to other similar types of algorithms. However, this algorithm suffers from diversity and local optima problems. To overcome these problems, we are proposing an improved version of the CSO algorithm by using opposition-based learning and the Cauchy mutation operator. We applied the opposition-based learning method to enhance the diversity of the CSO algorithm and we used the Cauchy mutation operator to prevent the CSO algorithm from trapping in local optima. The performance of our proposed algorithm was tested with several artificial and real datasets and compared with existing methods like K-means, particle swarm optimization, and CSO. The experimental results show the applicability of our proposed method.

Existing simulators for performance analysis of resource management techniques in edge computing have a limitation: they lack horizontal management features such as inter-server clustering and container registry placement. To address this issue, this paper proposes EdgeNet, a new simulator specialized for modeling of network overhead and server clustering algorithm in edge computing environments. EdgeNet provides a Python library that can be used to develop leader election algorithms for clustered server groups, facilitating research into horizontal resource management approaches that were previously difficult to study.

비지도 방식의 클러스터링은 주로 각 데이터에 대한 유사도나 거리에 기반하여 수행되며, 기본적으로 NP (Non-deterministic polynomial) Hard의 영역으로 알려져 있다. 각 노드에 대한 계산은 매트릭스에 기반하여 수행되는데, 노드의 수가 많은 경우 다른 노드와 비교하는 계산 시간이 매우 많이 소요될 수 밖에 없으며, 이를 단시간 내에 계산하기 위해서는 동적프로그래밍과 같은 컴퓨터 알고리즘이 수반되어야 한다. 이러한 계산 복잡도와 구현의 어려움으로 인해 빅데이터의 클러스터링은 유클리드나 코사인 유사도 등 몇 가지 전통적인 컴퓨터 거리 계산 방식에 국한되어 적용되고 그 계산 방식을 제공하는 주요한 클러스터링 라이브러리에 종속적으로 의존되어 왔다. 따라서 이러한 보편적인 클러스터링으로 계산이 불가능한 특수한 데이터의 경우에는 적용이 아예 불가능하거나 어려운 점이 존재할 수 있다. 예를 들어 개인별 직무 경력과 같은 데이터는, 특정인의 경력 정보가 다른 인력의 경력 정보와 비교를 할 수 있는데, 이를 어떻게 비교를 하여 그 거리를 특정화하고, 여러가지 “career pathway”를 분류해내고 검토하기 위한 특수한 클러스터링 알고리즘이 요구된다. 본 연구에서는 IT 분야의 경력정보 데이터를 활용하여 생명공학 분야에서 DNA 시퀀스에 대한 분류를 위해 활용되는 Optimal Matching 알고리즘을 활용하여 경력 정보의 계산한 후 이를 활용하여 클러스터링하는 시스템을 소개한다.

신체영역무선통신망(WBAN)에서 클러스터 헤드(CH) 선출 및 최적경로의 라우팅은 에너지 효율 향상과 네트워크 노드 운영수명 연장을 위해 해결해야 할 이슈이다. 이러한 연구를 위해 본 논문에서는 BKOA알고리즘과 그리드 기반 멀티홉 라우팅 프레임워크를 결합한 하이브리드 BKOA-GRID를 제안한다. 시뮬레이션 수행결과 제안된 BKOA-GRID는 PSO, LEACH, EEUC 등 기존 알고리즘보다 노드생존율 90%, 잔류에너지 지속성은 총 에너지의 약 60%를 유지하여 높은 에너지 효율을 보였다.

Cluster head(CH) election and optimal path routing in a Wireless Body Area Network(WBAN) are issues that must be addressed to improve energy efficiency and extend the operating life of network nodes. To address these issues, this paper proposes a hybrid BKOA-GRID (Black Kite Optimization Algorithm-GRID) framework, which integrates the Black Kite Optimization Algorithm with a grid-based multi-hop routing structure. Simulation results demonstrate that the proposed BKOA-GRID exhibits superior energy efficiency compared to existing algorithms such as PSO, LEACH, and EEUC, maintaining a node survival rate of 90% and preserving approximately 60% of the total residual energy.

무인항공기의 충돌 회피 기술은 장애물에 대한 탐지 기술과 충돌 여부 판단 및 회피 기술이 요구된다. 본 논문은 공선상에 존재하는 다중 정적 장애물에 대한 무인항공기의 충돌 회피를 수행하기 위하여, LiDAR를 활용한 장애물 탐지 알고리즘과 최근접점 기반의 충돌 인식 및 회피 알고리즘을 제안한다. 장애물 탐지를 수행하기 위하여 LiDAR의 측정 데이터 중 지면을 제거하는 전처리를 수행하고, K-평균 군집화 알고리즘을 활용하여 전처리된 데이터에서 장애물을 탐지 및 분류한다. 또한, 상대 항법을 통해 탐지한 다중 장애물의 절대 위치를 추정하며, 저주파 통과 필터를 활용하여 추정 위치를 보정한다. 탐지한 다중 정적 장애물과의 충돌 회피를 수행하기 위해 최근접점 기반의 충돌 인식 및 회피 알고리즘을 활용한다. 각 장애물 간의 거리를 활용하여 회피해야 하는 장애물 정보를 갱신하고, 갱신된 장애물 정보를 통해 충돌 인식 및 회피를 수행한다. 마지막으로 Gazebo 시뮬레이션 환경에서의 장애물 위치 추정, 충돌 인식 및 회피 결과 분석을 통해, 충돌 회피가 정상적으로 수행되는 것을 검증하였다.

Obstacle detection, collision recognition, and avoidance technologies are required the collision avoidance technology for UAVs. In this paper, considering collinear multiple static obstacle, we propose an obstacle detection algorithm using LiDAR and a collision recognition and avoidance algorithm based on CPA. Preprocessing is performed to remove the ground from the LiDAR measurement data before obstacle detection. And we detect and classify obstacles in the preprocessed data using the K-means clustering algorithm. Also, we estimate the absolute positions of detected obstacles using relative navigation and correct the estimated positions using a low-pass filter. For collision avoidance with the detected multiple static obstacle, we use a collision recognition and avoidance algorithm based on CPA. Information of obstacles to be avoided is updated using distance between each obstacle, and collision recognition and avoidance are performed through the updated obstacles information. Finally, through obstacle location estimation, collision recognition, and collision avoidance result analysis in the Gazebo simulation environment, we verified that collision avoidance is performed successfully.

본 논문은 공간객체의 영향력을 고려한 클러스터링을 위한 알고리즘인 DBSCAN-SI를 제안한다. DBSCAN-SI는 기존의 DBSCAN과 DBSCAN-W를 확장한 것으로 공간클러스터링 시 비공간 속성들을 영향력으로 변환한다. DBSCAN-SI는 클러스터링에 사용되는 속성에 의한 영향력이 클수록 클러스터에 포함될 확률을 높여주어, 단지 공간적인 거리뿐만이 아니라 영향력의 크기를 반영하여 군집화를 수행하기 위한 알고리즘이다. 이 논문에서 제안한 클러스터링 기법은 주변에 있는 객체들이 특정 속성 중심으로 보았을 때, 영향력이 큰 객체임에도 불구하고 주변에 객체가 드물게 있으므로 인하여 클러스터에서 배제되게 되는 기존 알고리즘의 단점을 보완해 줄 수 있다.

This paper proposes DBSCAN-SI that is an algorithm for clustering with influences of spatial objects. DBSCAN-SI that is extended from existing DBSCAN and DBSCAN-W converts from non-spatial properties to the influences of spatial objects during the spatial clustering. It increases probability of inclusion to the cluster according to the higher the influences that is affected by the properties used in clustering and executes the clustering not only respect the spatial distances, but also volume of influences. For the perspective of specific property-centered, the clustering technique proposed in this paper can makeup the disadvantage of existing algorithms that exclude the objects in spite of high influences from cluster by means of being scarcely close objects around the cluster.

Among the optimization method, GA (genetic algorithm) is a very powerful searching method enough to compete with design sensitivity analysis method. GA is very easy to apply, since it dose not require any design sensitivity information. However, GA has been computationally not efficient due to huge repetitive computation. In this study, parallel computation is adopted to improve computational efficiency. Paralleled GA is introduced on a clustered LINUX based personal computer system.