• The KIPS Transactions:PartB
• /
• v.14B no.3 s.113
• /
• pp.199-214
• /
• 2007

Recently, there have been a lot of works trying to realize automated composition of semantic web services though application of AI planning techniques. The traditional AI planning techniques, however, have some limitations: it is not easy to represent a web service process with complex control constructs as an action or a plan; it is hardly possible to consider enough the rich information contained in domain ontologies during the planning process; it is impossible to model directly the data flow from the outputs of a web service to the inputs of another web service; it is difficult to predict and deal with uncertainty and dynamics of the environment because the plan generation phase is supposed to be separated from the plan execution phase. In order to overcome some of these limitations, this paper suggests a reactive planning approach to automated composition of semantic web services. Through some experiments using several e-commerce web services, we found that the reactive planning is an effective way to realize automated composition of semantic web services.

• Computers and Concrete
• /
• v.24 no.3
• /
• pp.271-282
• /
• 2019

Concrete undergoes a series of thermo-based physio-chemical changes once exposed to elevated temperatures. Such changes adversely alter the composition of concrete and oftentimes lead to fire-induced explosive spalling. Spalling is a multidimensional, complex and most of all sophisticated phenomenon with the potential to cause significant damage to fire-exposed concrete structures. Despite past and recent research efforts, we continue to be short of a systematic methodology that is able of accurately assessing the tendency of concrete to spall under fire conditions. In order to bridge this knowledge gap, this study explores integrating novel artificial intelligence (AI) techniques; namely, artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS) and genetic algorithm (GA), together with traditional statistical analysis (multilinear regression (MLR)), to arrive at state-of-the-art procedures to predict occurrence of fire-induced spalling. Through a comprehensive datadriven examination of actual fire tests, this study demonstrates that AI techniques provide attractive tools capable of predicting fire-induced spalling phenomenon with high precision.

• The Journal of Information Systems
• /
• v.30 no.3
• /
• pp.177-199
• /
• 2021

Purpose This study explores the development direction of the artificial intelligence curriculum as a universal education that enhances the ability of college students to flexibly use artificial intelligence these days, where artificial intelligence education is spreading, and the educational components based on this are subdivided according to the characteristics of each major. Design/methodology/approach In order to develop the educational purpose of the subject and the detailed educational curriculum suitable for the subject of education, we first analyzed domestic and foreign prior research related to artificial intelligence liberal arts education. As the main components derived by experts, the basic concept of artificial intelligence converges to literacy to read and write for everyday problem solving, as well as problem-solving ability to manipulate real data and software. Findings The results showed that In the artificial intelligence literacy module, trends and prospects of artificial intelligence and necessary competencies were checked, and cases applied to major fields were examined. In the AI utilization and application part, basic data analysis items and content composition were composed through creative thinking, logical thinking, and intelligence. In order to design the curriculum, a software development language suitable for each major area was first selected, and AI education content areas, elements, and packages were defined and designed for each major area to meet the objectives of the subject.

• Journal of Industrial Convergence
• /
• v.22 no.1
• /
• pp.177-183
• /
• 2024

The study investigates the impact of data quality on the performance of artificial intelligence (AI). To this end, the impact of labeling error levels on the performance of artificial intelligence was compared and analyzed through simulation, taking into account the similarity of data features and the imbalance of class composition. As a result, data with high similarity between characteristic variables were found to be more sensitive to labeling accuracy than data with low similarity between characteristic variables. It was observed that artificial intelligence accuracy tended to decrease rapidly as class imbalance increased. This will serve as the fundamental data for evaluating the quality criteria and conducting related research on artificial intelligence learning data.

• Journal of Digital Convergence
• /
• v.18 no.6
• /
• pp.9-16
• /
• 2020

The purpose of this study is to understand the public needs for AI education actively promoted and supported by the current government. In doing so, 11 metropolitan news articles and Twitter posts regarding AI education that have been posted from January 1, 2018 to December 31, 2019 were collected. Then, word frequency analysis using TF(Term Frequency) method and LDA(Latent Dirichlet Allocation) method of topic modeling analysis were conducted. The topics of the news articles turn out to be a macroscopic policy support such as 'training female manpower in the AI field' and 'curriculum reform of university and K-12', whereas the topics of twitter delineate more detailed social perception on future society, such as future competencies and pedagogical methods, including 'coexistence with intelligent robots', 'coding education', and 'humane education competence development'. The findings are expected to be used to suggest the implications for the composition and management of AI curriculum as well as the basic framework of human resources development in the future industry.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.5
• /
• pp.817-826
• /
• 2022

AI technology is being successfully introduced in many fields, and models deployed as a service are deployed with black box environment that does not expose the model's information to protect intellectual property rights and data. In a black box environment, attackers try to steal data or parameters used during training by using model output. This paper proposes a method of inferring the type of model to directly find out the composition of layer of the target model, based on the fact that there is no attack to infer the information about the type of model from the deep learning model. With ResNet, VGGNet, AlexNet, and simple convolutional neural network models trained with MNIST datasets, we show that the types of models can be inferred using the output values in the gray box and black box environments of the each model. In addition, we inferred the type of model with approximately 83% accuracy in the black box environment if we train the big and small relationship feature that proposed in this paper together, the results show that the model type can be infrerred even in situations where only partial information is given to attackers, not raw probability vectors.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.3
• /
• pp.89-100
• /
• 2023

Recently, on-device artificial intelligence (AI) solutions using mobile devices and embedded edge devices have emerged in various fields, such as computer vision, to address network traffic burdens, low-energy operations, and security problems. Although vision transformer deep learning models have outperformed conventional convolutional neural network (CNN) models in computer vision, they require more computations and parameters than CNN models. Thus, they are not directly applicable to embedded edge devices with limited hardware resources. Many researchers have proposed various model compression methods or lightweight architectures for vision transformers; however, there are only a few studies evaluating the effects of model compression techniques of vision transformers on performance. Regarding this problem, this paper presents a performance evaluation of vision transformers on embedded platforms. We investigated the behaviors of three vision transformers: DeiT, LeViT, and MobileViT. Each model performance was evaluated by accuracy and inference time on edge devices using the ImageNet dataset. We assessed the effects of the quantization method applied to the models on latency enhancement and accuracy degradation by profiling the proportion of response time occupied by major operations. In addition, we evaluated the performance of each model on GPU and EdgeTPU-based edge devices. In our experimental results, LeViT showed the best performance in CPU-based edge devices, and DeiT-small showed the highest performance improvement in GPU-based edge devices. In addition, only MobileViT models showed performance improvement on EdgeTPU. Summarizing the analysis results through profiling, the degree of performance improvement of each vision transformer model was highly dependent on the proportion of parts that could be optimized in the target edge device. In summary, to apply vision transformers to on-device AI solutions, either proper operation composition and optimizations specific to target edge devices must be considered.

• Nutrition Research and Practice
• /
• v.16 no.sup1
• /
• pp.47-56
• /
• 2022

This paper examines the process and evidence used to create the Dietary Reference Intake (DRI) of alpha-linolenic acid (ALA) and eicosapentaenoic acid (EPA) + docosahexaenoic acid (DHA) for Koreans. ALA (18:3n3) is an essential fatty acid, and EPA and DHA are known to have beneficial effects on cardiovascular disease risk and reduction of triglyceride levels. Various international organizations have suggested dietary recommendations for n-3 polyunsaturated fatty acids (PUFAs), including ALA, EPA, and DHA. A DRI for Koreans was established for the first time in 2020, specifically for the adequate intake (AI) of ALA and EPA + DHA. This recommendation was based on the average intake of ALA and EPA + DHA from the Korea National Health and Nutrition Examination Survey 2013-2017. For Korean infants, the AI of ALA and DHA was based on the fatty acid composition of maternal milk. Estimated average requirement and a tolerable upper intake level have not been set for n-3 PUFA due to insufficient evidence. In addition, the intake level of n-3 PUFA for prevention of chronic disease has also not been determined. Future studies and randomized controlled trials are required to establish the UL and to define the level for disease prevention.

• Journal of Ocean Engineering and Technology
• /
• v.33 no.5
• /
• pp.427-434
• /
• 2019

Artificial intelligence (AI)-aided research currently enjoys active use in a wide array of fields thanks to the rapid development of computing capability and the use of Big Data. Until now, forecasting methods were primarily based on physics models and statistical studies. Today, AI is utilized in disaster prevention forecasts by studying the relationships between physical factors and their characteristics. Current studies also involve combining AI and physics models to supplement the strengths and weaknesses of each aspect. However, prior to these studies, an optimization algorithm for the AI model should be developed and its applicability should be studied. This study aimed to improve the forecast performance by constructing a model for neural network optimization. An artificial neural network (ANN) followed the ever-changing path of a typhoon to produce similar typhoon predictions, while the optimization achieved by the neural network algorithm was examined by evaluating the activation function, hidden layer composition, and dropouts. A learning and test dataset was constructed from the available digital data of one typhoon that affected Korea throughout the record period (1951-2018). As a result of neural network optimization, assessments showed a higher degree of forecast accuracy.

• Journal of the Korean Institute of Gas
• /
• v.25 no.2
• /
• pp.64-71
• /
• 2021

Research on polymer matrix composites with excellent molding processability and mechanical properties in the automotive field including hydrogen fuel cell electric vehicles is expanding to Computer-Aided Engineering (CAE) to support the design of materials with specific mechanical properties. CAE automation requires the prediction of the mechanical properties and behavior of materials. Unlike single materials, the mechanical properties prediction of polymer matrix composites is difficult to explain with formulas because the mechanical behavior is complicated to be explained only by the relationship between the matrix and the filler. In this study, the stress-strain curve according to the composition of polymer matrix composites, which was difficult to predict due to its sensitivity to large plastic deformation and composition, was predicted based on machine learning of the test data. The developed model finds a complex correlation between matrix and filler types and compositions, and predicts the total stress-strain curve meaningfully even in the absence of learned test data. It is expected that the material design AI system can be completed in the future based on the developed model that predicts the mechanical properties of polymer matrix composites even for the combination and composition that have not been learned.