• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.4
• /
• pp.115-119
• /
• 2023

The natural language processing models used in current artificial intelligence are huge, causing various difficulties in processing and analyzing data in real time. In order to solve these difficulties, we proposed a method to improve the efficiency of processing by using less memory and checked the performance of the proposed model. The technique applied in this paper to evaluate the performance of the proposed model is to divide the large corpus by adjusting the number of attention heads and embedding size of the BERT[1] model to be small, and the results are calculated by averaging the output values of each forward. In this process, a random offset was assigned to the sentences at every epoch to provide diversity in the input data. The model was then fine-tuned for classification. We found that the split processing model was about 12% less accurate than the unsplit model, but the number of parameters in the model was reduced by 56%.

• The korean journal of orthodontics
• /
• v.53 no.5
• /
• pp.328-335
• /
• 2023

Objective: The aim of this study was to determine whether an association between study quality, other study characteristics, and Altmetric Attention Scores (AASs) existed in orthodontic studies. Methods: The Scopus database was searched to identify orthodontic studies published between January 1, 2017, and December 31, 2019. Articles that satisfied the eligibility criteria were included in this study. Study characteristics, including study quality were extracted and entered into a pre-pilot data collection sheet. Descriptive statistics were calculated. On an exploratory basis, random forest and gradient boosting machine learning algorithms were used to examine the influence of article characteristics on AAS. Results: In total, 586 studies with an AAS were analyzed. Overall, the mean AAS of the samples was 5. Twitter was the most popular social media platform for publicizing studies, accounting for 53.7%. In terms of study quality, only 19.1% of the studies were rated as having a high level of quality, with 41.8% of the studies deemed moderate quality. The type of social media platform, number of citations, impact factor, and study type were among the most influential characteristics of AAS in both models. In contrast, study quality was one of the least influential characteristics on the AAS. Conclusions: Social media platforms contributed the most to the AAS for orthodontic studies, whereas study quality had little impact on the AAS.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.9
• /
• pp.93-104
• /
• 2023

In modern society, the service method is tended to prefer the non-face-to-face method rather than the face-to-face method. However, services that recommend products such as shoes will inevitably be face-to-face method. In this paper, for the purpose of non-face-to-face service, a system that a foot size is automatically measured and some shoes are recommended based on the measurement result is proposed. To analyze the performance of the proposed method, size measurement error rate and recommendation performance were analyzed. In the recommendation performance experiments, a total of 10 methods for similarity calculation were used and the recommendation method with the best performance among them was applied to the system. From the experiments, the error rate the foot size was small and the recommendation performance was possible to derive significant results. The proposed method is at the laboratory level and needs to be expanded and applied to the real environment. Also, the recommendation method considering design could be needed in the future work.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.129-132
• /
• 2022

Super-resolution is the process of converting a low-quality image into a high-quality image. This study was conducted using ESPCN. In a super-resolution deep neural network, different quality images can be output even when receiving the same input data according to the activation function that determines the weight when passing through each node. Therefore, the purpose of this study is to find the most suitable activation function for super-resolution by applying the activation functions ReLU, ELU, and Swish and compare the quality of the output image for the same input images. The CelebaA Dataset was used as the dataset. Images were cut into a square during the pre-processing process then the image quality was lowered. The degraded image was used as the input image and the original image was used for evaluation. As a result, ELU and swish took a long time to train compared to ReLU, which is mainly used for machine learning but showed better performance.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.535-538
• /
• 2022

For Plug & Play of IoT devices, we develop a module that recognizes the type of USB, which is a typical wired interface of IoT devices, through image recognition. In order to drive an IoT device, a driver for communication and device hardware is required. The wired interface for connecting to the IoT device is recognized by using the image obtained through the camera of smartphone shooting to recognize the corresponding communication interface. For USB, which is a most popular wired interface, types of USB are classified through artificial neural network-based machine learning. In order to secure sufficient data set of artificial neural networks, USB images are collected through the Internet, and additional image data sets are secured through image processing. In addition to the convolution neural networks, recognizers are implemented with various deep artificial neural networks, and their performance is compared and evaluated.

• Geomechanics and Engineering
• /
• v.30 no.6
• /
• pp.489-502
• /
• 2022

In this study, a Gaussian process regression (GPR) model as well as six GPR-based metaheuristic optimization models, including GPR-PSO, GPR-GWO, GPR-MVO, GPR-MFO, GPR-SCA, and GPR-SSO, were developed to predict fly-rock distance in the blasting operation of open pit mines. These models included GPR-SCA, GPR-SSO, GPR-MVO, and GPR. In the models that were obtained from the Soungun copper mine in Iran, a total of 300 datasets were used. These datasets included six input parameters and one output parameter (fly-rock). In order to conduct the assessment of the prediction outcomes, many statistical evaluation indices were used. In the end, it was determined that the performance prediction of the ML models to predict the fly-rock from high to low is GPR-PSO, GPR-GWO, GPR-MVO, GPR-MFO, GPR-SCA, GPR-SSO, and GPR with ranking scores of 66, 60, 54, 46, 43, 38, and 30 (for 5-fold method), respectively. These scores correspond in conclusion, the GPR-PSO model generated the most accurate findings, hence it was suggested that this model be used to forecast the fly-rock. In addition, the mutual information test, also known as MIT, was used in order to investigate the influence that each input parameter had on the fly-rock. In the end, it was determined that the stemming (T) parameter was the most effective of all the parameters on the fly-rock.

• International Journal of Computer Science & Network Security
• /
• v.23 no.1
• /
• pp.147-152
• /
• 2023

Mostly in motor fault detection the instantaneous values 3 axis vibration and 3phase current in time domain are acquired and converted to frequency domain. Vibrations are more useful in diagnosing the mechanical faults and motor current has remained more useful in electrical fault diagnosis. With having some experience and knowledge on the behavior of acquired data the electrical and mechanical faults are diagnosed through signal processing techniques or combine machine learning and signal processing techniques. In this paper, a single-layer LSTM based condition monitoring system is proposed in which the instantaneous values of three phased motor current are firstly acquired in simulated motor in in health and supply imbalance conditions in each of three stator currents. The acquired three phase current in time domain is then used to train a LSTM network, which can identify the type of fault in electrical supply of motor and phase in which the fault has occurred. Experimental results shows that the proposed single layer LSTM algorithm can identify the electrical supply faults and phase of fault with an average accuracy of 88% based on the three phase stator current as raw data without any processing or feature extraction.

• Informatization Policy
• /
• v.30 no.1
• /
• pp.67-88
• /
• 2023

The expansion of the scope of product liability is necessary because the industrial environment has changed following the enactment of the Product Liability Act. Unlike human-coded algorithms, artificial intelligence is black-boxed according to machine learning, and even developers cannot explain the results. In particular, since the cause of the problem by artificial intelligence is unknown, the responsibility is unclear, and compensation for victims is not easy. This is because software or artificial intelligence is a non-object, and its productivity is not recognized under the Product Liability Act, which is limited to movable property. As a desperate measure, productivity may be recognized if it is stored or embedded in the medium. However, it is not reasonable to apply differently depending on the medium. The EU revise the product liability guidelines that recognize product liability when artificial intelligence is included. Although compensation for victims is the value pursued by the Product Liability Act, the essence has been overlooked by focusing on productivity. Even if an accident occurs using an artificial intelligence-adopted service, however, it is desirable to present standards according to practical risks instead of unconditionally holding product responsibility.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.2
• /
• pp.313-320
• /
• 2023

The cumulative registration status of dogs is from 1.07 million in 2016 to 2.32 million in 2020. Animal registration is increasing by more than 10% every year, and accordingly, a name must be decided when registering a dog. We want to give a name that fits the characteristics of a dog's appearance, but there are many difficulties in naming it. This paper explains the development of a system for recognizing dog images and recommends dog names based on similar objects or food. This system extracts similarities with dogs' images through models that learn images of various objects and foods, and recommends dog names based on similarities. In addition, by recommending additional related words based on the image data of the result value, it was possible to provide users with various options, increase convenience, and increase interest and fun. Through this system, it is expected that users will be able to solve their concerns about naming their dogs, check names that suit their dogs comfortably, and give them various options through various recommended names to increase satisfaction.

• Journal of Internet Computing and Services
• /
• v.24 no.2
• /
• pp.37-45
• /
• 2023

In line with the 4th Industrial Revolution, facility maintenance technologies for building smart factories are receiving attention and are being advanced. In addition, technology is being applied to smart farms and smart fisheries following smart factories. Among them, in the case of a recirculating aquaculture system, there is a motor pump that circulates water for a stable quality environment in the tank. Motor pump maintenance activities for recirculating aquaculture system are carried out based on preventive maintenance and data obtained from vibration sensor. Preventive maintenance cannot cope with abnormalities that occur before prior planning, and vibration sensors are affected by the external environment. This paper proposes an anomaly detection algorithm that utilizes ADTK, a Python open source, for motor pump anomaly detection based on data collected through current sensors that are less affected by the external environment than noise, temperature and vibration sensors.