• Journal of the Korea Society of Computer and Information
• /
• v.24 no.12
• /
• pp.17-24
• /
• 2019

The biggest problem in the AI field, which has become a hot topic in recent years, is how to deal with the lack of training data. Since manual data construction takes a lot of time and efforts, it is non-trivial for an individual to easily build the necessary data. On the other hand, automatic data construction needs to handle data quality issue. In this paper, we introduce a method to automatically extract the data required to develop Korean speech command recognizer from the web and to automatically select the data that can be used for training data. In particular, we propose a modified ResNet model that shows modest performance for the automatically constructed Korean speech command data. We conducted an experiment to show the applicability of the command set of the health and daily life domain. In a series of experiments using only automatically constructed data, the accuracy of the health domain was 89.5% in ResNet15 and 82% in ResNet8 in the daily lives domain, respectively.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.11
• /
• pp.549-554
• /
• 2016

ConceptNet is a common sense knowledge base which is formed in a semantic graph whose nodes represent concepts and edges show relationships between concepts. As it is difficult to make knowledge base integrity, a knowledge base often suffers from incompleteness problem. Therefore the quality of reasoning performed over such knowledge bases is sometimes unreliable. This work presents neural tensor networks which can alleviate the problem of knowledge bases incompleteness by reasoning new assertions and adding them into ConceptNet. The neural tensor networks are trained with a collection of assertions extracted from ConceptNet. The input of the networks is two concepts, and the output is the confidence score, telling how possible the connection between two concepts is under a specified relationship. The neural tensor networks can expand the usefulness of ConceptNet by increasing the degree of nodes. The accuracy of the neural tensor networks is 87.7% on testing data set. Also the neural tensor networks can predict a new assertion which does not exist in ConceptNet with an accuracy 85.01%.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.5
• /
• pp.1431-1445
• /
• 2022

We construct a deep neural network model named ECGResNet. This model can diagnosis diseases based on 12-lead ECG data of eight common cardiovascular diseases with a high accuracy. We chose the 16 Blocks of ResNet50 as the main body of the model and added the Squeeze-and-Excitation module to learn the data information between channels adaptively. We modified the first convolutional layer of ResNet50 which has a convolutional kernel of 7 to a superposition of convolutional kernels of 8 and 16 as our feature extraction method. This way allows the model to focus on the overall trend of the ECG signal while also noticing subtle changes. The model further improves the accuracy of cardiovascular and cerebrovascular disease classification by using a fully connected layer that integrates factors such as gender and age. The ECGResNet model adds Dropout layers to both the residual block and SE module of ResNet50, further avoiding the phenomenon of model overfitting. The model was eventually trained using a five-fold cross-validation and Flooding training method, with an accuracy of 95% on the test set and an F1-score of 0.841.We design a new deep neural network, innovate a multi-scale feature extraction method, and apply the SE module to extract features of ECG data.

• Journal of the Korean Society of Radiology
• /
• v.18 no.2
• /
• pp.93-99
• /
• 2024

The digital medical imaging, especially, computed tomography (CT), should necessarily be considered in terms of noise distribution caused by converting to X-ray photon to digital imaging signal. Recently, the denoising technique based on deep learning architecture is increasingly used in the medical imaging field. Here, we evaluated noise reduction effect according to various noise types based on the U-net deep learning model in the lung CT images. The input data for deep learning was generated by applying Gaussian noise, Poisson noise, salt and pepper noise and speckle noise from the ground truth (GT) image. In particular, two types of Gaussian noise input data were applied with standard deviation values of 30 and 50. There are applied hyper-parameters, which were Adam as optimizer function, 100 as epochs, and 0.0001 as learning rate, respectively. To analyze the quantitative values, the mean square error (MSE), the peak signal to noise ratio (PSNR) and coefficient of variation (COV) were calculated. According to the results, it was confirmed that the U-net model was effective for noise reduction all of the set conditions in this study. Especially, it showed the best performance in Gaussian noise.

• The KIPS Transactions:PartB
• /
• v.15B no.4
• /
• pp.323-330
• /
• 2008

Since complex diseases are caused by interactions of multiple genes, traditional statistical methods are limited in its power to predict the onset of a complex disease. Recently new approaches using machine learning techniques are introduced. Neural nets are a suitable model to find patterns in complex data. When large amount of data are fed into a neural net, however, it takes a long time for learning and finding patterns. In this study we suggest a new model that combines the set association, which is a statistical technique to find important SNPs associated with complex diseases, and neural network. We experiment with SNP data related to asthma to test the effectiveness of our model. Our model shows higher prediction accuracy and shorter execution time than neural net only. We expect our model can be used effectively to predict the onset of other complex diseases.

• Korean Journal of Ichthyology
• /
• v.18 no.3
• /
• pp.223-233
• /
• 2006

The annual and spatial changes in the species composition of catch off Yeosu were analyzed using the daily sales slip catch data by a set net in the inshore waters off Dolsan Islands in Yeosu from April to October 2001, off Yeon Islands of Yeosu from April to October 2002 and in the offshore waters off Dolsan Islands of Yeosu from April to December 2003, respectively. Scomberomorus niphonius, Seriola spp., Trichiurus lepturus, Engraulis japonicus, Sarda orientalis, Todarodes pacificus, Pampus echinogaster, Sardinella zunasi, Scomber japonicus, Lophius litulon and Loligo beka were dominant species in abundance, indicating that pelagic fish were mainly caught by a set net off Yeosu. S. zunasi, P. echinogaster, Platycephalus indicus and L. beka inhabited mainly in the inshore waters, and S. niphonius, Seriola spp., T. lepturus, P. echinogaster, T. pacificus, Takifugu porphyreus and Pagrus major resided mainly in the offshore waters as the pelagic resident species. E. japonicus was a representative dominant species moving between the inshore and the offshore waters seasonally. S. zunasi and E. japonicus occurred in the inshore waters, and E. japonicus, L. litulon and Seriola spp. begain to be caught in the deep offshore waters in spring. Total catch was high during the summer season by migration of the open sea species such as T. lepturus, S. niphonius, S. japonicus, Seriola spp., S. orientalis, P. echinogaster and T. pacificus. In fall, S. niphonius, E. japonicus, Sphyraena pinguis, Siganus fuscescens and Leiognathus nuchalis were dominant in the inshore waters, and S. niphonius, P. echinogaster, Hyporhamphus sajori, S. japonicus and T. lepturus continued to occur from summer in the offshore waters but total catch decreased, indicating the typical seasonal variation pattern of the temperate region. Most of catchable fishes by a set net were the pelagic species showing a significant temporal variation. Collection and analysis of daily catch data by large set nets can be used to determine seasonal variation in species composition of pelagic fish in a study area.

• KIISE Transactions on Computing Practices
• /
• v.24 no.1
• /
• pp.53-57
• /
• 2018

A poisoning accident is often caused by a situation in which people eat poisonous mushrooms because they cannot distinguish between edible mushrooms and poisonous mushrooms. In this paper, we propose an automatic mushroom recognition system by using the convolutional neural network. We collected 1478 mushroom images of 38 species using image crawling, and used the dataset for learning the convolutional neural network. A comparison experiment using AlexNet, VGGNet, and GoogLeNet was performed using the collected datasets, and a comparison experiment using a class number expansion and a fine-tuning technique for transfer learning were performed. As a result of our experiment, we achieve 82.63% top-1 accuracy and 96.84% top-5 accuracy on test set of our dataset.

• Journal of the Korean Society for Heat Treatment
• /
• v.22 no.1
• /
• pp.16-22
• /
• 2009

The rolling process is an efficient and economical approach for the manufacturing of plate metals. In the rolling process, the temperature variation is very critical for plate thickness accuracy. The main cause of thickness variation in hot plate mills is the non-uniform temperature distribution along the length of the slab. Also the exit plate thickness is mainly affected by the rolling conditions such as mill modulus, plate thickness and plate width. Hence the thickness variation in top-end is also dependent on these factors. Therefore this study has concentrated on determining the correct amounts of thickness variation due to top-end temperature drop and process parameters.

• Asia pacific journal of information systems
• /
• v.7 no.1
• /
• pp.99-111
• /
• 1997

This paper presents an application of an artificial neural net to the implementation of decision class analysis (DCA), together with the generation of a decision model influence diagram. The diagram is well-known as a good tool for knowledge representation of complex decision problems. Generating influence diagram model is known to in practice require much time and effort, and the resulting model can be generally applicable to only a specific decision problem. In order to reduce the burden of modeling decision problems, the concept of DCA is introduced. DCA treats a set of decision problems having some degree of similarityz as a single unit. We propose a method utilizing a feedforward neural net with supervised learning rule to develop DCA based on influence diagram, which method consists of two phases: Phase l is to search for relevant chance and value nodes of an individual influence diagram from given decision and specific situations and Phase II elicits arcs among the nodes in the diagram. We also examine the results of neural net simulation with an example of a class of decision problems.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.38 no.2
• /
• pp.172-178
• /
• 2002

This study presents a dynamic simulation of a purse seine net behavior Mathematical model suitable for purse seining, which is based on data from a series of previous simulations, various field experiments, is modelized as a set of mass-spring system. In this model, a number of meshes are approximated as one mass point, each of which connected to its neighbors by massless springs, the equations of motion are derived from considering internal force from the springs and external forces such as resistance and gravitation. This simulation shows the quantitative state on every mass point of the net and purse line during the shooting and pursing phases. So it is possible that performance of a purse seine net be analyzed using various and evolving parameters such as the shooting speed, the hauling speed, the size or type of the sinker, float and twine, also the hanging ratio etc.