• Journal of Internet Computing and Services
• /
• v.24 no.4
• /
• pp.1-13
• /
• 2023

As federated learning brings a large paradigm to modern artificial intelligence research, efforts are being made to incorporate federated learning into research in various fields. However, researchers who apply federated learning face the problem of choosing a federated learning framework and benchmark tool suitable for their situation and purpose. This study aims to present guidelines for selection of federated learning frameworks and benchmark tools considering the circumstances of researchers who apply federated learning in practice. In particular, there are three main contributions in this study. First, it generalizes the situation of the researcher applying federated learning by combining it with the goal of federated learning and proposes guidelines for selecting a federated learning framework suitable for each situation. Second, it shows the suitability of selection by comparing the characteristics and performance of each federated learning framework to the researcher. Finally, the limitations of the existing federated learning framework and a plan for real-world federated learning operation are proposed.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.4
• /
• pp.959-968
• /
• 1999

several papers have reported that parallel heuristics or hybrid approaches combining several heuristics can get better results. However, the parallelization and hybridization of any search methods on the single CPU type computer need enormous computation time. that case, we need more elegant combination method. For this purpose, we propose Fast Model of Parallel Heuristics(FMPH). FMPH is based on the island model of parallel genetic algorithms and takes local search to the elite solution obtained form each island(sub group). In this paper we introduce how can we adapt FMPH to the job-shop scheduling problem notorious as the most difficult NP-hard problem and report the excellent results of several famous benchmark problems.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.15 no.6
• /
• pp.257-266
• /
• 2015

There is no polynomial-time algorithm that can be obtain the optimal solution for economic load dispatch problem with non-convex fuel cost functions. Therefore, electrical field uses quadratic fuel cost function unavoidably. This paper proposes a valve-point optimization (VPO) algorithm for economic load dispatch problem with non-convex fuel cost functions. This algorithm sets the initial values to maximum powers $P_i{\leftarrow}P_i^{max}$ for each generator. It then reduces the generation power of generator i with an average power cost of $_{max}\bar{c}_i$ to a valve point power $P_{ik}$. The proposed algorithm has been found to perform better than the extant heuristic methods when applied to 13 and 40-generator benchmark data. This paper consequently proves that the optimal solution to economic load dispatch problem with non-convex fuel cost functions converges to the valve-point power of each generator.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.23 no.8
• /
• pp.2032-2041
• /
• 1998

In this work, we propose a novel bit allocation method that is to minimize overall distortions subject ot the bit rate constraint. We partition the original bitallocation problem into 'macroblock level bit allocation' problems that can be solved by conventional Lagrangian mutiplier methods and a 'frame level bit allocation' problem. To tackle the frame level problem, 'two-phase optimization' algorithm is used with iter-frame dependency model. While the existing approaches are almost impossible to find the macroblock-unit result for the moving picture coding system due to high computational complexity, the proposed algorithm can drastically reduce the computational loads by the problem partitioning and can obtain the result close to the optimal solution. Because the optimally allocated results can be used as a benchmark for bit allocation methods, the upper performance limit, or a basis for approximation method development, we expect that the proposed algorithm can be very useful for the bit allocation related works.

• Smart Structures and Systems
• /
• v.17 no.6
• /
• pp.957-980
• /
• 2016

Structural damage detection (SDD) is a challenging task in the field of structural health monitoring (SHM). As an exploring attempt to the SDD problem, a hybrid self-adaptive Firefly-Nelder-Mead (SA-FNM) algorithm is proposed for the SDD problem in this study. First of all, the basic principle of firefly algorithm (FA) is introduced. The Nelder-Mead (NM) algorithm is incorporated into FA for improving the local searching ability. A new strategy for exchanging the information in the firefly group is introduced into the SA-FNM for reducing the computation cost. A random walk strategy for the best firefly and a self-adaptive control strategy of three key parameters, such as light absorption, randomization parameter and critical distance, are proposed for preferably balancing the exploitation and exploration ability of the SA-FNM. The computing performance of the SA-FNM is evaluated and compared with the basic FA by three benchmark functions. Secondly, the SDD problem is mathematically converted into a constrained optimization problem, which is then hopefully solved by the SA-FNM algorithm. A multi-step method is proposed for finding the minimum fitness with a big probability. In order to assess the accuracy and the feasibility of the proposed method, a two-storey rigid frame structure without considering the finite element model (FEM) error and a steel beam with considering the model error are taken examples for numerical simulations. Finally, a series of experimental studies on damage detection of a steel beam with four damage patterns are performed in laboratory. The illustrated results show that the proposed method can accurately identify the structural damage. Some valuable conclusions are made and related issues are discussed as well.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.5
• /
• pp.15-24
• /
• 2008

For high-dimensional classification tasks, such as face recognition, the number of samples is smaller than the dimensionality of the samples. In such cases, a problem encountered in linear discriminant analysis-based methods for dimension reduction is what is known as the small sample size (SSS) problem. Recently, to solve the SSS problem, a way of employing a dissimilarity-based classification(DBC) has been investigated. In DBC, an object is represented based on the dissimilarity measures among representatives extracted from training samples instead of the feature vector itself. In this paper, we propose a new method of optimizing DBCs using multi-level fusion strategies(MFS), in which fusion strategies are employed to represent features as well as to design classifiers. Our experimental results for benchmark face databases demonstrate that the proposed scheme achieves further improved classification accuracies.

• Journal of information and communication convergence engineering
• /
• v.1 no.3
• /
• pp.104-108
• /
• 2003

A biologically inspired fuzzy multilayer perceptron is proposed in this paper. The proposed algorithm is established under consideration of biological neuronal structure as well as fuzzy logic operation. We applied this suggested learning algorithm to benchmark problem in neural network such as exclusive OR and 3-bit parity, and to digit image recognition problems. For the comparison between the existing and proposed neural networks, the convergence speed is measured. The result of our simulation indicates that the convergence speed of the proposed learning algorithm is much faster than that of conventional backpropagation algorithm. Furthermore, in the image recognition task, the recognition rate of our learning algorithm is higher than of conventional backpropagation algorithm.

• Transactions on Control, Automation and Systems Engineering
• /
• v.4 no.2
• /
• pp.109-113
• /
• 2002

Van de Vusse reactor is known as a highly nonlinear chemical process and has been considered by a number of researchers as a benchmark problem for nonlinear chemical process. Various identification methods for nonlinear system are also verified by applying these methods to Van de Vusse reactor. From the point of view of identification, only the Volterra kernel of second order has been obtained until now. In this paper, the authors show that Volterra kernels of nonlinear Van de Vusse reactor of up to 3rd order are obtained by use of M-sequence correlation method. A pseudo-random M-sequence is applied to Van de Vusse reactor as an input and its output is measured. Taking the crosscorrelation function between the input and the output, we obtain up to 3rd order Volterra kernels, which is the highest order Volterra kernel obtained until now for Van de Vusse reactor. Computer simulations show that when Van de Vusse chemical process is identified by use of up to 3rd order Volterra kernels, a good agreement is observed between the calculated output and the actual output.

• Industrial Engineering and Management Systems
• /
• v.12 no.4
• /
• pp.330-335
• /
• 2013

A surface mount device (SMD) assembles electronic components on printed circuit boards (PCB). Since a component assembly process is a bottleneck process in a PCB assembly line, making an efficient SMD plan is critical in increasing the PCB assembly line productivity. Feeder assignment is an important part of the SMD plan optimization. In this paper, we propose a feeder re-assign improvement algorithm for a specific type of SMD machine with a piano type multi-head gantry. Computational results on some real-world benchmark data sets show the effectiveness of our proposed algorithm.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.26 no.9
• /
• pp.1319-1324
• /
• 2002

A theoretical model on shock-vortex interaction is investigated using a numerical technique to solve Navier-Stokes equations. The shock-vortex interaction generated by this model based on the classical Rankin vortex is precisely investigated for a benchmark problem: Dosanjh and Weeks experiment. In terms of shock dynamics, the interaction is categorized to three stages: shock distortion, shock split, and shock-shock interaction. The quadrupolar structure of the sound source produced by the interaction is far supported with the present model, and the difference between experiment and theoretical model is also discussed in this paper.