• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.1
• /
• pp.85-92
• /
• 2020

Deep learning shows outstanding performance in image and video analysis, such as object classification, object detection and semantic segmentation. In this paper, it is analyzed that the performances of deep learning models can be affected by characteristics of train dataset. It is proposed as a method for selecting activation function and optimization algorithm of deep learning to classify facial expression. Classification performances are compared and analyzed by applying various algorithms of each component of deep learning model for CK+, MMI, and KDEF datasets. As results of simulation, it is shown that genetic algorithm can be an effective solution for optimizing components of deep learning model.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2007.06a
• /
• pp.466-469
• /
• 2007

최근 늘어가는 통신 서비스에 대한 관심은 관련 이론 및 기술의 발전을 주도해 왔다. 이러한 통신 시스템에 관련한 문제들 중에 가장 근간이 되는 문제 중 하나는 최적 통신 걸침 나무 (OCST: Optimal Communication Spanning Tree) 문제이다. 본 논문에서는 이러한 최적 통신 걸침 나무 문제에 대한 해법으로 최적 통신 걸침 나무 문제에 적합한 새로운 유전자 표현법을 이용한 유전 알고리즘을 설계하여 구현한다. 제안하는 유전 알고리즘은 n개의 노드에 대해 n-2개의 숫자 열로 표현가능하다고 알려진 기존의 $Pr\ddot{u}fer$수 기반의 유전자 표현법을 개선하여 보다 더 최적 통신 걸침 문제에 적합하도록 고안한 새로운 유전자 표현법을 이용한다. 임의로 생성된 예제에 대한 수치 실험을 통해 통신시스템의 기본 문제 중 하나인 최적 통신 걸침 문제의 해법으로서의 제안 알고리즘의 유용성과 효율성을 확인한다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.8 no.5
• /
• pp.197-203
• /
• 2008

In this work, we present a novel genetic approach to solve the problem of combining power control and data rate transmission adjustment for the performance enhancement of the next generation CDMA system. We obtained the optimal solution of multi rate and power control problem by compromising slightly on SIR limit values. The proposed algorithm was able to handle many more users with comparable or faster convergent service. While this paper considered two kinds of fitness function such as maximizing the total transmission data rate and maximizing the acceptable mobiles of CDMA cellular network, the evaluation function combining these two cases or others can be also easily implemented. The simulation results showed the effectiveness and validity of our approach.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.5
• /
• pp.131-136
• /
• 2017

Recently, with the development of Internet media, many document materials have become exponentially increasing on the web. These materials are described, and the information on what is the most by this text are classified according. However, the text has meant that many have room for ambiguous interpretation must look at it from various angles in order to interpret them correctly. In conventional classification methods it was simply a classification only have the appearance of the text. In this paper, we analyze it in terms genetic algorithm and local preserving based techniques and implemented a clustering system fragmentation them. Finally, the performance of this paper was evaluated based on the implementation results compared to traditional methods.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.8 no.11
• /
• pp.1665-1674
• /
• 2013

In this paper, we propose a genetic algorithm(GA) based locomotion method of a quadruped robot with waist joint, which makes a quadruped robot walk on the slop efficiently. In the proposed method, we first derive the kinematic model of a quadruped robot with waist joint and then set the gene and the fitness function for GA. In addition, we determine the best attitude for a quadruped robot and the landing point of a foot in the walk space, which has the optimal energy stability margin(ESM). Finally, we verify the effectiveness of the proposed method by comparing with the performance of the previous method through the computer simulations.

• Development and Reproduction
• /
• v.25 no.2
• /
• pp.113-121
• /
• 2021

This article aims to introduce German discussion on the approval of the non-invasive prenatal diagnosis (NIPD), which started with the development of PrenaTest^® by LifeCodexx AG. The discussion started with the concern that the non-invasive nature of NIPD, such as PrenaTest^®, may rapidly expand the use and scope of similar tests, thus leading to a new era of eugenics. Based on this concern, the need for clear clinical guidelines on specific indications for NIPD has been suggested. Along the same line, it was discussed whether PrenaTest^® is against the Basic Law prohibiting discrimination on grounds of disability and whether the test is outside the scope of the purpose of gene testing limited by Genetic Diagnosis Act. Through such discussion, the Federal Ministry of Health of Germany established the preconditions for inclusion of NIPD in the German public health insurance system. For this, the German motherhood guideline was amended and the information for the insured persons provided to pregnant women was included in the amended guideline. Such discussion made in Germany provides insight on which points should be considered when various gene testings are accepted in Korea, in which genetic communication has not been systematized yet. In particular, German counseling system for pregnant women will provide valuable insights for Korea where the direction for regulations on abortion has not been established even after the ruling by the Constitutional Court that charges for abortion are against the constitution.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.3
• /
• pp.63-67
• /
• 2024

Recently, artificial intelligence (AI) workloads encompassing various industries such as smart logistics, FinTech, and entertainment are being executed on the cloud. In this paper, we address the scheduling issues of various AI workloads on a multi-tenant cloud system composed of heterogeneous GPU clusters. Traditional scheduling decreases GPU utilization in such environments, degrading system performance significantly. To resolve these issues, we present a new scheduling approach utilizing genetic algorithm-based optimization techniques, implemented within a process-based event simulation framework. Trace driven simulations with diverse AI workload traces collected from Alibaba's MLaaS cluster demonstrate that the proposed scheduling improves GPU utilization compared to conventional scheduling significantly.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.9
• /
• pp.1216-1223
• /
• 2018

As field of application of AMI, AMR uses the power line as the primary means of communication. PLC has a big merit without installation of the new network for communication in a field using the power line which is the existing equipment. However, there is a serious obstacle in commercialization for the instability by noise and communication environment. Therefore, the technical method for maintaining the communication state which overcome such demerit and was stabilized is required essentially. PLC routing technology is applied with the alternative plan now. The routing technology currently managed by field includes many problems by applying the algorithm of an elementary level. PLC routing path search problem can be modeled with the problem of searching for optimal solution as similar to such as optimal routing problem and TSP(Travelling salesman problem). In this paper, in order to search for a PLC routing path efficiently and to choose the optimal path, GA(Genetic Algorithm) was applied. Although PLC was similar in optimal solution search as compared with typical GA, it also has a difference point by the characteristic of communication, and presented the new methodology over this. Moreover, the validity of application technology was verified by showing the experimental result to which GA is applied and analyzing as compared with the existing algorithm.

• ETRI Journal
• /
• v.27 no.3
• /
• pp.243-249
• /
• 2005

A mobile antenna for multimedia communications with Ku-band geostationary satellite KOREASAT-3 and JSAT-2A is presented. The forward link of the satellite communication is 11.7 to 12.75 GHz, and the return link is 14.0 to 14.5 GHz. The mobile antenna is designed to be a stair structure using 24 active phased array elements in order to provide a low profile, and to be at a non-periodic array distance using the genetic algorithm. Also, the designed antenna uses the double beam forming method for stable satellite tracking. The fabricated mobile antenna is examined using various experiments to confirm its capability for practical application. From the measured results, the fabricated mobile antenna system is confirmed to have a good performance.

• Journal of electromagnetic engineering and science
• /
• v.9 no.4
• /
• pp.223-228
• /
• 2009

In this paper, a quadrafilar antenna is designed to have a lower profile as an essential part of the size reduction technique and lower back radiation(i.e. higher forward radiation) for the S-band telemetry/telecommand(TM/TC) function of a communication satellite. Particularly, to meet the challenging requirements on the higher isolation between the TM/TC antennas and simultaneously a smaller size, the lowest back radiation and lowest cross-polarization, the optimal physical dimensions of the quadrafilar antenna are found by using the Genetic Algorithm(GA). To prove the validity of the proposed antenna design, its 3D electromagnetic analysis and measured results are compared, showing good agreement.