• Journal of the Korea Society of Computer and Information
• /
• v.28 no.4
• /
• pp.31-39
• /
• 2023

In this paper, we propose an Internal/External Knowledge Distillation (IEKD), which utilizes both external correlations between feature maps of heterogeneous models and internal correlations between feature maps of the same model for transferring knowledge from a teacher model to a student model. To achieve this, we transform feature maps into a sequence format and extract new feature maps suitable for knowledge distillation by considering internal and external correlations through a transformer. We can learn both internal and external correlations by distilling the extracted feature maps and improve the accuracy of the student model by utilizing the extracted feature maps with feature matching. To demonstrate the effectiveness of our proposed knowledge distillation method, we achieved 76.23% Top-1 image classification accuracy on the CIFAR-100 dataset with the "ResNet-32×4/VGG-8" teacher and student combination and outperformed the state-of-the-art KD methods.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.05a
• /
• pp.505-508
• /
• 2016

In city, tracing of power transmission lines is difficult due to compound installation of overhead and underground lines, transposition, bad view caused by trees or big buildings. It is hard problem for electrical technician on site to trace power transformers or power lines to serve customers in 3 phase -4 wires power distribution systems. It is necessary that the correct and fast tracing method is required for load balancing among distribution lines. Old technology use to trace lines with high power impulse injection. Our proposed method uses to trace lines with very small power high frequency signal injection. Simulation models for 3-phase power transformers, 3-phase wire lines, and customer loads are described to investigate the transmission characteristics of high frequency power line carrier. Distribution lines have only a limited ability to carry higher frequencies. Typically power transformers in the distribution system prevent propagating the higher frequency carrier signal. The proposed method uses the limited propagation ability to identify the power transformer to serve customers. The system consists of a transmitter and a receiver with power-line communication module. Some experiments are conducted to verify the theoretical concepts in a big commercial building. Also some simulations are done to help and understand the concepts by using MATLAB Simulink simulator.

• Smart Media Journal
• /
• v.11 no.3
• /
• pp.9-17
• /
• 2022

Algorithmic trading, which uses algorithms to trade financial products, has a problem in that the results are not stable due to many factors in the market. To alleviate this problem, ensemble techniques that combine trading algorithms have been proposed. However, there are several problems with this ensemble method. First, the trading algorithm may not be selected so as to satisfy the minimum performance requirement (more than random) of the algorithm included in the ensemble, which is a necessary requirement of the ensemble. Second, there is no guarantee that an ensemble model that performed well in the past will perform well in the future. In order to solve these problems, a method for selecting trading algorithms included in the ensemble model is proposed as follows. Based on past data, we measure the contribution of the trading algorithms included in the ensemble models with high performance. However, for contributions based only on this historical data, since there are not enough past data and the uncertainty of the past data is not reflected, the contribution distribution is approximated using the Dirichlet distribution, and the contribution values are sampled from the contribution distribution to reflect the uncertainty. Based on the contribution distribution of the trading algorithm obtained from the past data, the Transformer is trained to predict the future contribution. Trading algorithms with high predicted future contribution are selected and included in the ensemble model. Through experiments, it was proved that the proposed ensemble method showed superior performance compared to the existing ensemble methods.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.3
• /
• pp.359-366
• /
• 2022

In this paper, we present the Korean menu-ordering Sentence Text-to-Speech (TTS) system using conformer-based FastSpeech2. Conformer is the convolution-augmented transformer, which was originally proposed in Speech Recognition. Combining two different structures, the Conformer extracts better local and global features. It comprises two half Feed Forward module at the front and the end, sandwiching the Multi-Head Self-Attention module and Convolution module. We introduce the Conformer in Korean TTS, as we know it works well in Korean Speech Recognition. For comparison between transformer-based TTS model and Conformer-based one, we train FastSpeech2 and Conformer-based FastSpeech2. We collected a phoneme-balanced data set and used this for training our models. This corpus comprises not only general conversation, but also menu-ordering conversation consisting mainly of loanwords. This data set is the solution to the current Korean TTS model's degradation in loanwords. As a result of generating a synthesized sound using ParallelWave Gan, the Conformer-based FastSpeech2 achieved superior performance of MOS 4.04. We confirm that the model performance improved when the same structure was changed from transformer to Conformer in the Korean TTS.

• Smart Media Journal
• /
• v.12 no.1
• /
• pp.9-16
• /
• 2023

Recently, a convolutional neural network (CNN) based system is being developed to overcome the limitations of human resources in the apple quality classification of farmhouse. However, since convolutional neural networks receive only images of the same size, preprocessing such as sampling may be required, and in the case of oversampling, information loss of the original image such as image quality degradation and blurring occurs. In this paper, in order to minimize the above problem, to generate a image patch based graph of an original image and propose a random walk-based positional encoding method to apply the graph transformer model. The above method continuously learns the position embedding information of patches which don't have a positional information based on the random walk algorithm, and finds the optimal graph structure by aggregating useful node information through the self-attention technique of graph transformer model. Therefore, it is robust and shows good performance even in a new graph structure of random node order and an arbitrary graph structure according to the location of an object in an image. As a result, when experimented with 5 apple quality datasets, the learning accuracy was higher than other GNN models by a minimum of 1.3% to a maximum of 4.7%, and the number of parameters was 3.59M, which was about 15% less than the 23.52M of the ResNet18 model. Therefore, it shows fast reasoning speed according to the reduction of the amount of computation and proves the effect.

• The Journal of Society for e-Business Studies
• /
• v.27 no.1
• /
• pp.15-28
• /
• 2022

Due to recent expansion of online market such as clothing, utilizing customer review has become a major marketing measure. User review has been used as a tool of analyzing sentiment of customers. Sentiment analysis can be largely classified with machine learning-based and lexicon-based method. Machine learning-based method is a learning classification model referring review and labels. As research of sentiment analysis has been developed, multi-modal models learned by images and video data in reviews has been studied. Characteristics of words in reviews are differentiated depending on products' and customers' categories. In this paper, sentiment is analyzed via considering review data and metadata of products and users. Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), Self Attention-based Multi-head Attention models and Bidirectional Encoder Representation from Transformer (BERT) are used in this study. Same Multi-Layer Perceptron (MLP) model is used upon every products information. This paper suggests a multi-modal sentiment analysis model that simultaneously considers user reviews and product meta-information.

• Korean Journal of Remote Sensing
• /
• v.39 no.6_1
• /
• pp.1413-1425
• /
• 2023

The increasing frequency of wildfires due to climate change is causing extreme loss of life and property. They cause loss of vegetation and affect ecosystem changes depending on their intensity and occurrence. Ecosystem changes, in turn, affect wildfire occurrence, causing secondary damage. Thus, accurate estimation of the areas affected by wildfires is fundamental. Satellite remote sensing is used for forest fire detection because it can rapidly acquire topographic and meteorological information about the affected area after forest fires. In addition, deep learning algorithms such as convolutional neural networks (CNN) and transformer models show high performance for more accurate monitoring of fire-burnt regions. To date, the application of deep learning models has been limited, and there is a scarcity of reports providing quantitative performance evaluations for practical field utilization. Hence, this study emphasizes a comparative analysis, exploring performance enhancements achieved through both model selection and data design. This study examined deep learning models for detecting wildfire-damaged areas using Landsat 8 satellite images in California. Also, we conducted a comprehensive comparison and analysis of the detection performance of multiple models, such as U-Net and High-Resolution Network-Object Contextual Representation (HRNet-OCR). Wildfire-related spectral indices such as normalized difference vegetation index (NDVI) and normalized burn ratio (NBR) were used as input channels for the deep learning models to reflect the degree of vegetation cover and surface moisture content. As a result, the mean intersection over union (mIoU) was 0.831 for U-Net and 0.848 for HRNet-OCR, showing high segmentation performance. The inclusion of spectral indices alongside the base wavelength bands resulted in increased metric values for all combinations, affirming that the augmentation of input data with spectral indices contributes to the refinement of pixels. This study can be applied to other satellite images to build a recovery strategy for fire-burnt areas.

• International Journal of Computer Science & Network Security
• /
• v.21 no.9
• /
• pp.281-291
• /
• 2021

The Islamic jurisprudential legal system represents an essential component of the Islamic religion, that governs many aspects of Muslims' daily lives. This creates many questions that require interpretations by qualified specialists, or Muftis according to the main sources of legislation in Islam. The Islamic jurisprudence is usually classified into branches, according to which the questions can be categorized and classified. Such categorization has many applications in automated question-answering systems, and in manual systems in routing the questions to a specialized Mufti to answer specific topics. In this work we tackle the problem of automatic categorisation of Islamic jurisprudential legal questions using deep learning techniques. In this paper, we build a hierarchical deep learning model that first extracts the question text features at two levels: word and sentence representation, followed by a text classifier that acts upon the question representation. To evaluate our model, we build and release the largest publicly available dataset of Islamic questions and answers, along with their topics, for 52 topic categories. We evaluate different state-of-the art deep learning models, both for word and sentence embeddings, comparing recurrent and transformer-based techniques, and performing extensive ablation studies to show the effect of each model choice. Our hierarchical model is based on pre-trained models, taking advantage of the recent advancement of transfer learning techniques, focused on Arabic language.

• Journal of the Korean Society for Railway
• /
• v.9 no.6 s.37
• /
• pp.652-656
• /
• 2006

Constant load model is widely used for an electric train to perform the static analysis of AT (Auto Transformer) feeding systems. In this model, the train will be considered as a constant load model when it drives or as a constant source model when it applies regenerative brake. However there must be some constraints imposed on the pantagraph voltage in actual operations. These constraints are established for the reason of protecting the feeding facilities from excessive rise of regenerative braking voltage or guaranteeing the minimum traction power of train. In normal operating situation, the pantagraph voltage of the train should be maintained within these limits. Keeping these facts in minds, we suggest new methods or analyzing AT feeding systems using the constant power models with the conditions of voltage constraints. The simulation results from a sample system using the proposed method illustrate both the states of system variables and the supply-demand relation of power among the trains and the systems very clearly, so it is believed that the proposed method yields more accurate results than conventional methods do. The proposed methods are believed to contribute to the assessment of TCR-TSC for compensating reactive powers too.

• Progress in Superconductivity and Cryogenics
• /
• v.15 no.4
• /
• pp.10-14
• /
• 2013

We have clearly investigated with respect to the survival of the insulator at cryogenic temperature under the electrical stress. The breakdown and voltage-time characteristics of turn-to-turn models for point contact geometry and surface contact geometry using copper multi wrapped with polyimide film for an HTS transformer were investigated under AC and impulse voltage at 77 K. Polyimide film (Kapton) 0.025 mm thick is used for multi wrapping of the electrode. As expected, the breakdown voltages for the surface contact geometry are lower than that of the point contact geometry, because the contact area of the surface contact geometry is lager than that of the point contact geometry. The time to breakdown t50 decreases as the applied voltage is increased, and the lifetime indices increase slightly as the number of layers is increased. The electric field amplitude at the position where breakdown occurs is about 80 % of the maximum electric field value. The relationship between survival probability and the electrical stress at cryogenic temperature was evident.