• Journal of Power Electronics
• /
• v.2 no.2
• /
• pp.95-103
• /
• 2002

The paper deals with the control of the synchronous reluctance machine without position senser. A method for the computation of the transformation angle out of terminal voltages and currents is presented. The injection of test signals allows operation at zero speed. Fundamental for this control scheme is the angle estimation method over the whole operating range including field-weakening is discussed in detail. The implementation of the angle estimation method in a rotor-oriented control scheme and practical results are demonstrated.

• Transactions of Materials Processing
• /
• v.7 no.5
• /
• pp.474-480
• /
• 1998

In the present study a domain decomposition scheme using the substructuring method is developed for the computational efficiency of the finite element analysis of metal forming processes. in order to avoid calculation of an inverse matrix during the substructuring procedure, the modified Cholesky decomposition method is implemented. As obtaining the data independence by the substructuring method the program is easily paralleized using the Parallel Virtual machine(PVM) library on a work-station cluster connected on networks. A numerical example for a simple upsetting is calculated and the speed-up ratio with respect to various number of subdomains and number of processors. The efficiency of the parallel computation is discussed by comparing the results.

• Journal of Platform Technology
• /
• v.11 no.4
• /
• pp.71-79
• /
• 2023

In this paper, we analyze the characteristics of machine learning workloads and, based on them, propose a distributed in-memory caching technique to improve the performance of machine learning workloads. The core of machine learning workload is model training, and model training is a computationally intensive task. Performing machine learning workloads in a Kubernetes-based cloud environment in which the computing framework and storage are separated can effectively allocate resources, but delays can occur because IO must be performed through network communication. In this paper, we propose a distributed in-memory caching technique to improve the performance of machine learning workloads performed in such an environment. In particular, we propose a new method of precaching data required for machine learning workloads into the distributed in-memory cache by considering Kubflow pipelines, a Kubernetes-based machine learning pipeline management tool.

• The Transactions of the Korean Institute of Electrical Engineers B
• /
• v.52 no.10
• /
• pp.501-506
• /
• 2003

A Superconducting Rotary Machine (SRM) is characterized by an air-cored machine with its rotor iron and stator iron teeth removed. For this reason, the SRM is featured by 3D magnetic flux distribution, which decreases in the direction of axis. Therefore, 3D magnetic field analysis method is required to know about characteristic of magnetic field distribution of SRM. In this paper, 3D flux distribution of SRM is calculated by analytical method. The magnetic field distribution of the field coils is calculated by Biot-Savart equation. The magnetic core is represented by magnetic surface polarities. This paper describes the combined use of above methods for the total field computation, and compares results of analytical method and 3D FEM(Finite Element Method).

• Journal of Korean Institute of Industrial Engineers
• /
• v.27 no.2
• /
• pp.158-168
• /
• 2001

We present a tabu search (TS) algorithm to minimize maximum lateness on a single machine in the presence of sequence dependent setup times and dynamic job arrivals. The TS algorithm starts with a feasible schedule generated by a modified ATCS (Apparent Tardiness Cost with Setups) rule, then through a series of search steps it improves the initial schedule. Results of extensive computational experiments show that the TS algorithm significantly outperforms a well-known RHP heuristic by Ovacik and Uzsoy, both on the solutions quality and the computation time. The performance advantage is particularly pronounced when there is high competition among jobs for machine capacity.

• Journal of Korean Institute of Industrial Engineers
• /
• v.18 no.1
• /
• pp.91-104
• /
• 1992

Flowshop scheduling problem is known to be NP-complete. Since the optimization apporach like branch-and-bound is limited by exponentially growing computation time, many heuristic methods have been developed. Total tardiness is one of the criteria that the researchers have recently considered in flowshop scheduling. There, however, are few literatures which studied the general (m machine)-flowshop scheduling under the total tardiness criterion. In this paper, a heuristic scheduling method to minimize total tardiness at the (m machine, n job)-flowshop is presented. A heuristic value function is proposed to be used as a dispatching criterion in initial schedule generation. And the schedule improving procedure, by pairwise interchange of tardy job with the job right ahead of it, is introduced. Illustrative examles and simulated results are presented.

• Electronics and Telecommunications Trends
• /
• v.34 no.4
• /
• pp.32-42
• /
• 2019

The performance of machine learning algorithms significantly depends on how a configuration of hyperparameters is identified and how a neural network architecture is designed. However, this requires expert knowledge of relevant task domains and a prohibitive computation time. To optimize these two processes using minimal effort, many studies have investigated automated machine learning in recent years. This paper reviews the conventional random, grid, and Bayesian methods for hyperparameter optimization (HPO) and addresses its recent approaches, which speeds up the identification of the best set of hyperparameters. We further investigate existing neural architecture search (NAS) techniques based on evolutionary algorithms, reinforcement learning, and gradient derivatives and analyze their theoretical characteristics and performance results. Moreover, future research directions and challenges in HPO and NAS are described.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.2
• /
• pp.219-224
• /
• 2020

Binary classification has been broadly investigated in machine learning. In addition, binary classification can be easily extended to multi class problems. To successfully utilize machine learning methods for classification tasks, preprocessing and feature extraction steps are essential. These are important steps to improve their classification performances. In this paper, we propose a new learning method based on weighted least squares. In the weighted least squares, designing weights has a significant role. Due to this necessity, we also propose a new technique to obtain weights that can achieve feature transformation. Based on this weighting technique, we also propose a method to combine the learning and feature extraction processes together to perform both processes simultaneously in one step. The proposed method shows the promising performance on five UCI machine learning data sets.

• 한국HCI학회:학술대회논문집
• /
• 2009.02a
• /
• pp.685-689
• /
• 2009

Evolutionary Computation is not only widely used in optimization and machine learning, but also being applied in creating novel structures and entities. This paper proposes evolutionary topic maps that can suggest new and creative knowledge not easily producible by humans. Interactive evolutionary computation method is applied into topic maps in order to accept human evaluation on feasibility of intermediate topic maps. Evolutionary topic maps are creativity support tools, helping users to encounter new and creative knowledge. Further work can greatly improve the system by providing more operations, preventing over-convergence, and overcoming user fatigue problem by providing more intuitive user interface, better visualization, and interpolation mechanisms.

• Proceedings of the Korean Statistical Society Conference
• /
• 2004.11a
• /
• pp.65-66
• /
• 2004

The aim of this article is to illustrate what soft computing is and how important it is.