• Journal of the Korea Society of Computer and Information
• /
• v.29 no.3
• /
• pp.43-54
• /
• 2024

Recently, there has been a lot of research on utilizing Language Models, and Large Language Models have achieved innovative results in various tasks. However, the practical application faces limitations due to the constrained resources and costs required to utilize Large Language Models. Consequently, there has been recent attention towards methods to effectively utilize models within given resources. Curriculum Learning, a methodology that categorizes training data according to difficulty and learns sequentially, has been attracting attention, but it has the limitation that the method of measuring difficulty is complex or not universal. Therefore, in this study, we propose a methodology based on data heterogeneity-based Curriculum Learning that measures the difficulty of data using reliable prior information and facilitates easy utilization across various tasks. To evaluate the performance of the proposed methodology, experiments were conducted using 5,000 specialized documents in the field of information communication technology and 4,917 documents in the field of healthcare. The results confirm that the proposed methodology outperforms traditional fine-tuning in terms of classification accuracy in both LoRA fine-tuning and full fine-tuning.

• Journal of the Korean Institute of Gas
• /
• v.28 no.1
• /
• pp.85-99
• /
• 2024

As the proliferation of hydrogen electric vehicles accelerates, there is observed diversification in hydrogen refueling station models. This diversification raises safety concerns for different types of stations. This study conducted a quantitative risk assessment of a multi-vehicle hydrogen station, capable of simultaneously refueling cars, buses, and trucks. Utilizing Gexcon's Effects&Riskcurves Software, scenarios of fire and explosion due to hydrogen leaks were assessed. The study calculated the impact distances from radiative heat and explosion overpressure, and measured risks to nearby buildings and populations. The largest impact distance was from fires and explosions at dispensers and high-pressure storage units. High-pressure storage contributes most significantly to personal and societal risk. The study suggests that conservative safety distances and proper protective measures for these facilities can minimize human and material damage in the event of a hydrogen leak.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.2
• /
• pp.531-536
• /
• 2024

Control drones, which are recently classified as smart agricultural machines in the agricultural field, are striving to build smart control and automatic control systems by combining hardware and software in order to shorten working hours and increase the effectiveness of control in the aging era of rural areas. In this paper, the characteristics of the nozzle dedicated to the control drone were analyzed as a basic study for the establishment of management control and automatic control systems. In order to consider various variables such as the type of various drone models, controller, wind, flight speed, flight altitude, weather conditions, and UAV pesticide types, related studies are needed to be able to present the drug spraying criteria in consideration of the characteristics and versatility of the nozzle. Therefore, to enable the consideration of various variables, flow analysis (CFD) simulation was conducted based on the self-designed nozzle, and the theoretical and experimental values of the droplet distribution were compared and analyzed through water reduction experiments. In the future, we intend to calculate accurate scattering in consideration of various variables according to drone operation and use it in management control and automatic control systems.

• Journal of Practical Engineering Education
• /
• v.16 no.3_spc
• /
• pp.291-296
• /
• 2024

This study analyzed various machine learning models that predict employment outcomes after vocational training using pre-assessment data of disabled vocational trainees. The study selected and utilized the most appropriate machine learning models based on a data set containing various personal characteristics, including trainees' gender, age, and type of disability. Through this analysis, the goal is to improve the employment rate and job satisfaction of disabled trainees using only pre-assessment data. As a result, it presents a universal approach that can be applied not only to people with disabilities, but also to vocational trainees from a variety of backgrounds. This is expected to make an important contribution to the development and implementation of tailored vocational training programs, ultimately helping to achieve better employment outcomes and job satisfaction.

• Convergence Security Journal
• /
• v.23 no.4
• /
• pp.53-59
• /
• 2023

Various problems are occurring as teachers and student's parents communicate directly through mobile phones. In this study, a service model was proposed that allows teachers and student's parents to communicate through SNS platforms without knowing each other's mobile phone numbers. In the civil complaint communication service model proposed in this study provides, communication key sets are provided as public data, and a commonly used SNS platform uses the relevant relationship information to implement communication. This model also has expandability that can be applied not only to teachers, but also to ① officers who need to communicate with the parents of soldiers, ② nursing, health and nursing care personnel who frequently contact patient caregivers, and ③ welfare officials.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.34 no.1
• /
• pp.103-114
• /
• 2024

In a situation where drone-related crimes continue to rise, research in drone forensics becomes crucial for preventing and responding to incidents involving drones. Conducting forensic analysis on flight record files stored internally is essential for investigating illegal activities. However, analyzing flight record files generated through the exclusive DUML protocol requires a deep understanding of the protocol's structure and characteristics. Additionally, a forensic analysis tool capable of handling cryptographic payloads and analyzing various drone models is imperative. Therefore, this study presents the methods and characteristics of flight record files generated by drones. It also explains the structure of the flight record file and the features of the DUML packet. Ultimately, we conduct forensic analysis based on the presented structure of the DUML packet and propose an extension forensic analysis system that operates more universally than existing tools, performing expanded syntactic analysis.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.11 no.3
• /
• pp.176-184
• /
• 2007

The objective of this paper is to suggest a nonlinear analysis process to predict the structural behavior and strength of composite basement wall member combined with H-Pile. Therefore, the structural behavior of composite basement wall is studied and the special nonlinear characteristics of each elements such as H-Pile, concrete wall, and shear connectors are idealized using ATENA program. Finally, the result is compared with previous test result. Research result shows that there is a good co-relation between analysis and test results even if analysis result has little bit higher initial stiffness than test result. It can be concluded that the nonlinear behavior of composite basement wall is suitably predicted by using the contact element model in ATENA program as shear connector element.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.12 no.4
• /
• pp.149-160
• /
• 2008

In this study, the research was carried out on how to identify the modal parameter by acquiring the output response of the structure only through the smart sensor system. The objective of this research is to verify the performance and the on-site adaptability of the smart sensor system that have been actively researched as the advanced measuring system so far. Smart Sensor System was developed so that the real-time dynamic measurement can be performed by means of MEMS-type accelerated sensor, 8 bit CPU, wireless MODEM. In the modal parameter identification test, random excitation was added to the cantilever beam, and then the response of the structure was obtained using the smart sensor system and the wire measurement system respectively. In analyzing the data, modal parameter was identified using NExT & ERA algorithm. Furthermore, the optimal measurement location was selected through EOT algorithm in order to obtain the qualified output response. Result of the test, it was possible to verify the on-site applicability of the smart sensor.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.4
• /
• pp.607-612
• /
• 2024

Generative AI is a type of artificial intelligence technology that produces various types of data. With the success of ChatGPT, the generative AI market is blooming. As the generative AI market develops, generative AI is being applied in various industries. In this paper, we discuss the trends, applications, and directions for improvement. Currently, generative AI is trained on domain knowledge and data, and it is evolving towards Vertical AI. In the future, generative AI could be extended to AGI, which makes decisions and processes on its own like a human, to be used flexibly in various environments.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.8
• /
• pp.85-92
• /
• 2024

In this paper presents a comparative analysis of the performance between the latest object detection model, YOLOv10, and its previous versions. YOLOv10 introduces NMS-Free training, an enhanced model architecture, and an efficiency-centric design, resulting in outstanding performance. Experimental results using the COCO dataset demonstrate that YOLOv10-N maintains high accuracy of 39.5% and low latency of 1.84ms, despite having only 2.3M parameters and 6.7G floating-point operations (FLOPs). The key performance metrics used include the number of model parameters, FLOPs, average precision (AP), and latency. The analysis confirms the effectiveness of YOLOv10 as a real-time object detection model across various applications. Future research directions include testing on diverse datasets, further model optimization, and expanding application scenarios. These efforts aim to further enhance YOLOv10's versatility and efficiency.