• Journal of Korea Multimedia Society
• /
• v.23 no.7
• /
• pp.803-811
• /
• 2020

This paper proposes an image classification algorithm that transforms the number of convolution layers in the residual block of ResNet, CNN's representative method. The proposed method modified the structure of 34/50 layer of ResNet structure. First, we analyzed the performance of small and many convolution layers for the structure consisting of only shortcut and 3 × 3 convolution layers for 34 and 50 layers. And then the performance was analyzed in the case of small and many cases of convolutional layers for the bottleneck structure of 50 layers. By applying the results, the best classification method in the residual block was applied to construct a 34-layer simple structure and a 50-layer bottleneck image classification model. To evaluate the performance of the proposed image classification model, the results were analyzed by applying to the cifar10 dataset. The proposed 34-layer simple structure and 50-layer bottleneck showed improved performance over the ResNet-110 and Densnet-40 models.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.9
• /
• pp.3712-3729
• /
• 2020

Smoke detection is helpful for early fire detection. With its large coverage area and low cost, vision-based smoke detection technology is the main research direction of outdoor smoke detection. We propose a two-stage smoke detection method combined with block Deep Normalization and Convolutional Neural Network (DNCNN) and visual change image. In the first stage, each suspected smoke region is detected from each frame of the images by using block DNCNN. According to the physical characteristics of smoke diffusion, a concept of visual change image is put forward in this paper, which is constructed by the video motion change state of the suspected smoke regions, and can describe the physical diffusion characteristics of smoke in the time and space domains. In the second stage, the Support Vector Machine (SVM) classifier is used to classify the Histogram of Oriented Gradients (HOG) features of visual change images of the suspected smoke regions, in this way to reduce the false alarm caused by the smoke-like objects such as cloud and fog. Simulation experiments are carried out on two public datasets of smoke. Results show that the accuracy and recall rate of smoke detection are high, and the false alarm rate is much lower than that of other comparison methods.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.45 no.1
• /
• pp.1-9
• /
• 2022

The PoN (Proof of Nonce) distributed consensus algorithm basically uses a non-competitive consensus method that can guarantee an equal opportunity for all nodes to participate in the block generation process, and this method was expected to resolve the first trilemma of the blockchain, called the decentralization problem. However, the decentralization performance of the PoN distributed consensus algorithm can be greatly affected by the network transaction transmission delay characteristics of the nodes composing the block chain system. In particular, in the consensus process, differences in network node performance may significantly affect the composition of the congress and committee on a first-come, first-served basis. Therefore, in this paper, we presented a problem by analyzing the decentralization performance of the PoN distributed consensus algorithm, and suggested a fairness control algorithm using a learning-based probabilistic acceptance rule to improve it. In addition, we verified the superiority of the proposed algorithm by conducting a numerical experiment, while considering the block chain systems composed of various heterogeneous characteristic systems with different network transmission delay.

• Journal of the Korean Society of Earth Science Education
• /
• v.5 no.1
• /
• pp.105-113
• /
• 2012

The purpose of this study was to examine the effects of storytelling skill on science learning motivation and space perception ability. For this study the 5 grade, 2 class was divided into a research group and a comparative group. The class was pre-tested in order to ensure the same standard. The research group had the science class with storytelling skill, and the comparative group had the class with teacher centered lectures for 10 classes in 10 weeks. The storytelling skill was focused on finding stories, constellation searching, story deciding, story hero deciding, story composition, storytelling completion. To prove the effects of this study, science learning motivation was split up according to attention power, relation, confidence, and sense of satisfaction. Also, space perception ability consisted of two-dimensional rotation, 3 dimension rotations, reflection, three-dimensional searching, number of block, and figure type in pattern. The results of this study are as follows. First, using storytelling skill was effective in science learning motivation. Second, using storytelling skill was effective in space perception ability. Also, after using storytelling skill was good reaction by students. As a result, the elementary science class with storytelling skill had the effects of developing science learning motivation and space perception ability. it means the science class with storytelling skill has potential possibilities and value to develop science learning motivation and space perception ability.

• The Journal of Korean Institute of Information Technology
• /
• v.17 no.12
• /
• pp.125-134
• /
• 2019

As the limitations of the passive recognition domain, which is not guaranteed transparency of the operation process, AI technology has a vulnerability that depends on the data. Human error is inherent because raw data for artificial intelligence learning must be processed and inspected manually to secure data quality for the advancement of AI learning. In this study, we examine the necessity of learning data management before machine learning by analyzing inaccurate cases of AI learning data and cyber security attack method through the approach from cyber security perspective. In order to verify the learning data integrity, this paper presents the direction of data-preserving artificial intelligence system, a blockchain-based learning data environment model. The proposed method is expected to prevent the threats such as cyber attack and data corruption in providing and using data in the open network for data processing and raw data collection.

• Journal of Nutrition and Health
• /
• v.27 no.9
• /
• pp.901-909
• /
• 1994

The effects of sardine oil(high in eicosapentaenoic acid : EPA) and tuna oil(high in docosahexaenoic acid : DHA, also high in EPA) on fatty acid composition of brain and learning ability were evaluated in male weanling Sprague-Dawley rats and compared with the effects of corn oil and beef tallow. Animals assigned by randomized block design to one of the four experimental diet groups containing dietary lipid at 15%(w/w) level were given ad libitum for 7 weeks. Food intake and body weight gain of the fish oil groups were significantly lower than those of the corn oil and beef tallow groups. However, brain weights of the groups were not significantly different. In the brain fatty acid composition, the corn oil group showed high concentrations of n-6 fatty acids, the fish oil groups of n-3 fatty acids, and the beef tallow group of saturated fatty acids. Brain EPA and DHA contents of the fish oil groups showed significantly higher than the other groups while the brain ratio of saturated/monounsaturated/polyunsaturated fatty acid was controlled in a narrow range. In a maze test, the fish oil groups appeared to arrive at the goal faster than the corn oil and beef tallow groups. It explained that EPA in diets might efficiently convert to DHA resulting in DHA accumulation in brain tissue and might increase the learning performance as DHA did.

• The Journal of Korean Association of Computer Education
• /
• v.19 no.2
• /
• pp.1-10
• /
• 2016

This study is for developing a programming education course to improve pre-service teachers' pedagogical content knowledge(PCK) of programming education. A 40-hour training course was designed with App Inventor, a block-based mobile programming environment, and with problem-based learning method and project-based learning method. After the curriculum was adopted to 12 undergraduate students, the effect of education was tested with a programming PCK questionnaire. As a result, after a 20-hour problem-based learning class, overall score and teaching method score were enhanced significantly. After another 20-hour project-based learning class, content knowledge, teaching method, and curriculum score were improved.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1400-1407
• /
• 2019

In this paper, we propose a support vector machine (SVM) based gas classifier that can support real-time self-learning. The modified sequential minimal optimization (MSMO) algorithm is employed to train the proposed SVM. By using a shared structure for learning and classification, the proposed SVM reduced the hardware area by 35% compared to the existing architecture. Our system was implemented with 3,337 CLB (configurable logic block) LUTs (look-up table) with Xilinx Zynq UltraScale+ FPGA (field programmable gate array) and verified that it can operate at the clock frequency of 108MHz.

• Journal of The Korean Association of Information Education
• /
• v.19 no.4
• /
• pp.489-498
• /
• 2015

In this study, we suggested the contents, teaching and learning method, and assessment types of algorithm education in elementary schools. First, we suggested the algorithm education contents; the expression, understanding, flowcharts, structure, results, correction, and improvement of algorithm. Second, we showed the algorithm teaching and learning methods; algorithm in our daily life, the unplugged activity, block programming and tangible programming. Finally, we analyzed all missions of 'Hour of Code' in Code.org, and suggested the algorithm assessment 4 types, which includes selecting, filling, correcting, predicting of appropriate algorithm.

• International Journal of Computer Science & Network Security
• /
• v.22 no.3
• /
• pp.155-162
• /
• 2022

Over the last 10 years, there has been rapid growth in the use of Machine Learning (ML) techniques to automate the process of intrusion threat detection at a scale never imagined before. This has prompted researchers, software engineers, and network specialists to rethink the applications of machine ML techniques particularly in the area of cybersecurity. As a result there exists numerous research documentations on the use ML techniques to detect and block cyber-attacks. This article is a systematic review involving the identification of published scholarly articles as found on IEEE Explore and Scopus databases. The articles exclusively related to the use of machine learning in Intrusion Detection Systems (IDS). Methods, concepts, results, and conclusions as found in the texts are analyzed. A description on the process taken in the identification of the research articles included: First, an introduction to the topic which is followed by a methodology section. A table is used to list identified research articles in the form of title, authors, methodology, and key findings.