• KIPS Transactions on Computer and Communication Systems
• /
• v.10 no.7
• /
• pp.191-198
• /
• 2021

Since the size of training dataset become large and the model is getting deeper to achieve high accuracy in deep learning, the deep neural network training requires a lot of computation and it takes too much time with a single node. Therefore, distributed deep learning is proposed to reduce the training time by distributing computation across multiple nodes. In this study, we propose hybrid allreduce strategy that considers the characteristics of each layer and communication and computational overlapping technique for synchronization of distributed deep learning. Since the convolution layer has fewer parameters than the fully-connected layer as well as it is located at the upper, only short overlapping time is allowed. Thus, butterfly allreduce is used to synchronize the convolution layer. On the other hand, fully-connecter layer is synchronized using ring all-reduce. The empirical experiment results on PyTorch with our proposed scheme shows that the proposed method reduced the training time by up to 33% compared to the baseline PyTorch.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.1
• /
• pp.1-12
• /
• 2019

In this paper, it was found that most of the machinery facilities problems generated by clients could be prevented in advance by systematically managing the mechanical equipment production process of small and medium enterprises (SMEs) that produce machinery facilities. Major point of this process is to establish an operating system that corresponds to reality of facility manufacturers as it represents 63% of machinery facilities problems that occur in customers and is a task that needs to be solved most intensively. Technical issues account for 23% of machinery facilities problems occurring at the client's companies and should be approached from a long-term perspective as they are directly related to the technical capabilities of the manufacturers. Organizational problems account for 14% of machinery facilities problems occurring in customer companies, and can change depending on the relationship of members and the nature of the human being, such as morality and motivation. In addition, we propose the establishment of an Internet-based production process management platform for smooth and efficient transfer of information between customers and machinery facilities manufacturers.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.14 no.1
• /
• pp.111-118
• /
• 2019

For the security of a vast amount of information, it has been started to diagnose the site as a way of operating and managing the information owned by a company holding assets, to establish indexes to check the actual status and all kinds of standards to obtain security, and also to classify the information assets based on that. This has been extended to many different areas including policies to operate and manage information assets, services, the management of owned devices as physical assets, and also the management of logical assets for application software and platforms. Some of these information assets are already being operated in reality as new technology in new areas, for example, Internet of Things. Of course, a variety of electronic devices like Smart Home are being used in ordinary families, and unlike in the past, these devices generate a series of information life cycles such as accumulating and processing information. Moreover, as even distribution is now being realized, we are facing a task to secure the stability of information assets and also information that assets are holding. The purpose of this study is to suggest and apply standard security policy by moduling methods for information assets owned by companies and even families and obtain the enhancement of confidentiality as well as integrity.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.10
• /
• pp.199-208
• /
• 2021

In this paper, the effect of artificial intelligence convergence education program that provides STEAM education using machine learning platform on elementary school students' STEAM literacy and learning flow was analyzed. A homogeneous group of 44 elementary school 6th graders was divided into an experimental group and a control group. The control group received 10 lessons of general subject convergence class, and the experimental group received 10 lessons of STEAM-based artificial intelligence convergence education using Machine learning for Kids. To develop the artificial intelligence convergence education program, the goals, achievement standards, and content elements of the 2015 revised curriculum to select subjects and class contents is analyzed. As a result of the STEAM literacy test and the learning flow test, there was a significant difference between the experimental group and the control group. In particular, it can be confirmed that the coding environment in which the artificial intelligence function is expanded has a positive effect on learners' learning flow and STEAM literacy. Among the sub-elements of convergence talent literacy, significant differences were found in the areas of personal competence such as convergence and creativity. Among the sub-elements of learning flow, significant differences were found in the areas such as harmony of challenge and ability, clear goals, focus on tasks, and self-purposed experiences. If further expanded research is conducted in the future, it will be a basic research for more effective education for the future.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.5
• /
• pp.195-200
• /
• 2021

Due to the development of video sharing platforms, the amount of video uploads is exploding. Such videos often include various types of music, among which cover songs are included. In order to protect the copyright of music, an algorithm to find the original song of the cover song is essential. However, it is not easy to find the original song because the cover song is a modification of the composition, speed and overall structure of the original song. So far, there is no known effective algorithm for searching the original song of the cover song. In this paper, we propose an algorithm for searching the original song of the cover song using the inflection points of the melody line. Inflection points represent the characteristic points of change in the melody sequence. The proposed algorithm compares the original song and the cover song using the sequence of inflection points for the representative phrase of the original song. Since the characteristics of the representative phrase are used, even if the cover song is a song made by modifying the overall composition of the song, the algorithm's search performance is excellent. Also, since the proposed algorithm uses only the features of the inflection point sequence, the memory usage is very low. The efficiency of the algorithm was verified through performance evaluation.

• Journal of Platform Technology
• /
• v.9 no.3
• /
• pp.52-62
• /
• 2021

Quality control is critical at manufacturing sites and is key to predicting the risk of quality defect before manufacturing. However, the reliability of manual quality control methods is affected by human and physical limitations because manufacturing processes vary across industries. These limitations become particularly obvious in domain areas with numerous manufacturing processes, such as the manufacture of major nuclear equipment. This study proposed a novel method for predicting the risk of quality defects by using natural language processing and machine learning. In this study, production data collected over 6 years at a factory that manufactures main equipment that is installed in nuclear power plants were used. In the preprocessing stage of text data, a mapping method was applied to the word dictionary so that domain knowledge could be appropriately reflected, and a hybrid algorithm, which combined n-gram, Term Frequency-Inverse Document Frequency, and Singular Value Decomposition, was constructed for sentence vectorization. Next, in the experiment to classify the risky processes resulting in poor quality, k-fold cross-validation was applied to categorize cases from Unigram to cumulative Trigram. Furthermore, for achieving objective experimental results, Naive Bayes and Support Vector Machine were used as classification algorithms and the maximum accuracy and F1-score of 0.7685 and 0.8641, respectively, were achieved. Thus, the proposed method is effective. The performance of the proposed method were compared and with votes of field engineers, and the results revealed that the proposed method outperformed field engineers. Thus, the method can be implemented for quality control at manufacturing sites.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.23 no.4
• /
• pp.173-183
• /
• 2020

In the midst of the aging of underground facilities in urban areas and anxiety about road excavation safety accidents, the Ministry of Land, Infrastructure and Transport began to build Underground Geospatial Infomation Map from 2015 as part of the 「ground subsidence prevention measures」 and efficient use of underground spaces. So, the scope is spreading every year. The current Underground Geospatial Infomation Map information is web-based and is operated in a desktop environment, so it is true that there are some limitations in its use in a field environment such as an excavation construction site. The Underground Geospatial Infomation Map, built and operated in a web-based environment, is a large-scale 3D data. Therefore, in order to service by transmitting data to the field without delay, it is necessary to lighten the Underground Geospatial Infomation Map data. In addition, the current Underground Geospatial Infomation Map is not unified in data formats such as 3DS and COLLADA, and the coordinate system method is also different in relative coordinates and absolute coordinates. In this study, by analyzing domestic and overseas prior research and technical use cases, a mobile Underground Geospatial Infomation Map data format and a lightweight method were presented, and a technology development was conducted to create a mobile underground space integration map in the presented format. In addition, the weight reduction rate was tested by applying 3D data compression technology so that data can be transmitted quickly in the field, and technology was developed that can be used by decompressing 3D data compressed in the field. finally, it aims to supplement the technology experimentally developed in this study and conduct additional research to produce it as software that can be used in the excavation site and use it.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.2
• /
• pp.705-713
• /
• 2021

Because the cage assembly serves as the launch platform, an accurate aim is essential to ensure shooting accuracy for the target. On the other hand, the abnormal rotation of the cage due to the directional errors of the K-MLRS has continuously caused quality problems. The quality problem of weapon systems may have a negative impact on the military's power loss. In this study, improvement plans were derived by examining the defects and analyzing the directional errors of the K-MLRS launcher. In addition, all possible causes of directional errors were derived from the flow diagram for cage directionality. Based on the results, the defense design through the software program was intended to prevent the loss of direction. Through this study, the signal error of the resolver was improved by preventing unspecific signals in the data. Furthermore, the directional judgment method was improved to minimize the impact of data distortion. Lastly, directional storage and verification methods were improved so that data for the cage rotation direction would not be affected by errors. For the design improvement method, the reliability was verified through the system applicability. This study is expected to be a reference for failure analysis and design for similar weapon systems in the future.

• The Korean Journal of Archival Studies
• /
• no.71
• /
• pp.49-89
• /
• 2022

This study suggests the necessity and domestic application plan a national authority database that promotes an integrated access, richer search, and understanding of historical information sources and archival resources distributed among cultural heritage institutions through the "Social Networks and Archive Context" project case. As the SNAC project was transformed into an international cooperative organization led by NARA, it was possible to secure a sustainable operating system and realize cooperative authority control. In addition, SNAC authority records have the characteristics of providing richer contextual information about life and history and social and intellectual network information compared to libraries. Through case analysis, First, like SNAC, a cooperative body led by the National Archives and having joint ownership of the National Library of Korea should lead the development and expand the scope of participating institutions. Second, in the cooperative method, take a structure in which divisions are made for each field with special strengths, but the main decision-making is made through the administrative team in which the two organizations participate. Third, development of scalable open source software that can collect technical information in various formats when constructing authority data, designing with the structure and elements of archival authority records, designing functions to control the quality of authority records, and building user-friendly interfaces and the need for a platform design reflecting content elements.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.1
• /
• pp.51-58
• /
• 2022

Drivable area detection, one of the main functions of advanced driver assistance systems, means detecting an area where a vehicle can safely drive. The drivable area detection is closely related to the safety of the driver and it requires high accuracy with real-time operation. To satisfy these conditions, V-disparity-based method is widely used to detect a drivable area by calculating the road disparity value in each row of an image. However, the V-disparity-based method can falsely detect a non-road area as a road when the disparity value is not accurate or the disparity value of the object is equal to the disparity value of the road. In a road environment including vegetation, such as a highway and a country road, the vegetation area may be falsely detected as the drivable area because the disparity characteristics of the vegetation are similar to those of the road. Therefore, this paper proposes a drivable area detection method and hardware architecture with a high accuracy in road environments including vegetation areas by reducing the number of false detections caused by V-disparity characteristic. When 289 images provided by KITTI road dataset are used to evaluate the road detection performance of the proposed method, it shows an accuracy of 90.12% and a recall of 97.96%. In addition, when the proposed hardware architecture is implemented on the FPGA platform, it uses 8925 slice registers and 7066 slice LUTs.