• Magazine of the Korean Society of Agricultural Engineers
• /
• v.43 no.2
• /
• pp.69-77
• /
• 2001

An artificial neural network model was developed to analyze and forecast daily steamflow flow a small watershed. Error Back propagation neural networks (EBPN) of daily rainfall and runoff data were found to have a high performance in simulating stremflow. The model adopts a gradient descent method where the momentum and adaptive learning rate concepts were employed to minimize local minima value problems and speed up the convergence of EBP method. The number of hidden nodes was optimized using Bayesian information criterion. The resulting optimal EBPN model for forecasting daily streamflow consists of three rainfall and four runoff data (Model34), and the best number of the hidden nodes were found to be 13. The proposed model simulates the daily streamflow satisfactorily by comparison compared to the observed data at the HS#3 watershed of the Baran watershed project, which is 391.8 ha and has relatively steep topography and complex land use.

• International Journal of Advanced Culture Technology
• /
• v.11 no.4
• /
• pp.464-469
• /
• 2023

As modern society develops, the number of diseases caused by bad smells is increasing. As it can harm people's health, it is important to predict in advance the extent to which bad smells may occur, inform the public about this, and take preventive measures. In this paper, we propose a hybrid neural network structure of CNN and LSTM that can be used to detect or predict the occurrence of odors, which are most required in manufacturing or real life, using odor complex sensors. In addition, the proposed learning model uses a complex odor sensor to receive four types of data, including hydrogen sulfide, ammonia, benzene, and toluene, in real time, and applies this data to the inference model to detect and predict the odor state. The proposed model evaluated the prediction accuracy of the training model through performance indicators based on accuracy, and the evaluation results showed an average performance of more than 94%.

• Korean Journal of Remote Sensing
• /
• v.36 no.6_2
• /
• pp.1567-1577
• /
• 2020

Urban land cover classification is role in urban planning and management. So, it's important to improve classification accuracy on urban location. In this paper, machine learning model, Support Vector Machine (SVM) and Artificial Neural Network (ANN) are proposed for urban land cover classification based on high resolution satellite imagery (KOMPSAT-3A). Satellite image was trained based on 25 m rectangle grid to create training data, and training models used for classifying test area. During the validation process, we presented confusion matrix for each result with 250 Ground Truth Points (GTP). Of the four SVM kernels and the two activation functions ANN, the SVM Polynomial kernel model had the highest accuracy of 86%. In the process of comparing the SVM and ANN using GTP, the SVM model was more effective than the ANN model for KOMPSAT-3A classification. Among the four classes (building, road, vegetation, and bare-soil), building class showed the lowest classification accuracy due to the shadow caused by the high rise building.

• KIPS Transactions on Software and Data Engineering
• /
• v.7 no.2
• /
• pp.69-76
• /
• 2018

Referring expressions are natural language constructions used to identify particular objects within a scene. In this paper, we propose a new deep neural network model for referring expression comprehension. The proposed model finds out the region of the referred object in the given image by making use of the rich information about the referred object itself, the context object, and the relationship with the context object mentioned in the referring expression. In the proposed model, the object matching score and the relationship matching score are combined to compute the fitness score of each candidate region according to the structure of the referring expression sentence. Therefore, the proposed model consists of four different sub-networks: Language Representation Network(LRN), Object Matching Network (OMN), Relationship Matching Network(RMN), and Weighted Composition Network(WCN). We demonstrate that our model achieves state-of-the-art results for comprehension on three referring expression datasets.

• Health Policy and Management
• /
• v.10 no.2
• /
• pp.1-21
• /
• 2000

This study reviewde various parametic and nonparametic method for forexasting hospital closures in Korea. We compared multivariate discriminant analysis, multivartiate logistic regression, classfication and regression tree, and neural network method based on hit ratio of each model for forecasting hospital closure. Like other studies in the literture, neural metwork analysis showed highest average hit ratio. For policy and business purposes, we combined the four analytical method and constructed a foreasting model that can be easily used to predict the probabolity of hospital closure given financial information of a hospital.

• Water Engineering Research
• /
• v.4 no.2
• /
• pp.59-68
• /
• 2003

The applicabilities and validities of two methodologies fur the prediction of THM (trihalomethane) formation in a water pipeline system were proposed and discussed. One is the multiple regression technique and the other is an artificial neural network technique. There are many factors which influence water quality, especially THMs formations in water pipeline systems. In this study, the prediction models of THM formation in water pipeline systems are developed based on the independent variables proposed by American Water Works Association(AWWA). Multiple linear/nonlinear regression models are estimated and three layer feed-forward artificial neural networks have been used to predict the THM formation in a water pipeline system. Input parameters of the models consist of organic compounds measured in water pipeline systems such as TOC, DOC and UV254. Also, the reaction time to each measuring site along pipeline is used as input parameter calculated by a hydraulic analysis. Using these variables as model parameters, four models are developed. And the predicted results from the four developed models are compared statistically to the measured THMs data set. It is shown that the artificial neural network approaches are much superior to the conventional regression approaches and that the developed models by neural network can be used more efficiently and reproduce more accurately the THMs formation in water pipeline systems, than the conventional regression methods proposed by AWWA.

• Journal of Korea Multimedia Society
• /
• v.24 no.12
• /
• pp.1598-1605
• /
• 2021

Recently, convolutional neural networks (CNNs) have been widely used with excellent performance in various computer vision fields, including super-resolution (SR). However, CNN is computationally intensive and requires a lot of memory, making it difficult to apply to limited hardware resources such as mobile or Internet of Things devices. To solve these limitations, network lightening studies have been actively conducted to reduce the depth or size of pre-trained deep CNN models while maintaining their performance as much as possible. This paper aims to lighten the SR CNN model, SRGAN, using the knowledge distillation among network lightening technologies; thus, it proposes four techniques with different methods of transferring the knowledge of the teacher network to the student network and presents experiments to compare and analyze the performance of each technique. In our experimental results, it was confirmed through quantitative and qualitative evaluation indicators that student networks with knowledge transfer performed better than those without knowledge transfer, and among the four knowledge transfer techniques, the technique of conducting adversarial learning after transferring knowledge from the teacher generator to the student generator showed the best performance.

• Geomechanics and Engineering
• /
• v.33 no.6
• /
• pp.567-581
• /
• 2023

The prediction and achievement of a proper rock fragmentation size is the main challenge of blasting operations in surface mines. This is because an optimum size distribution can optimize the overall mine/plant economics. To this end, this study attempts to develop four improved artificial intelligence models to predict rock fragmentation through cascaded forward neural network (CFNN) and radial basis function neural network (RBFNN) models. In this regards, the CFNN was trained by the Levenberg-Marquardt algorithm (LMA) and Conjugate gradient backpropagation (CGP). Further, the RBFNN was optimized by the Dragonfly Algorithm (DA) and teaching-learning-based optimization (TLBO). For developing the models, the database required was collected from the Midouk copper mine, Iran. After modeling, the statistical functions were computed to check the accuracy of the models, and the root mean square errors (RMSEs) of CFNN-LMA, CFNN-CGP, RBFNN-DA, and RBFNN-TLBO were obtained as 1.0656, 1.9698, 2.2235, and 1.6216, respectively. Accordingly, CFNN-LMA, with the lowest RMSE, was determined as the model with the best prediction results among the four examined in this study.

• Journal of the Korea Society for Simulation
• /
• v.25 no.3
• /
• pp.1-14
• /
• 2016

This paper proposes four general wireless gateway types, which are distinguished by their own processing ways to connect between a wireless sub-network of things and the mobile internet that links mobile network to internet step by step. In this paper, we also design general processing procedures of these four types using the state transition model. Gateways of each types were developed on the basis of the resulted state transition models and their performances were evaluated through several tests, analyzed, and compared each other. As the results of our evaluation, compared with the other types, the type, which combines both of a low-power Sleep-interrupt way and polling ways for receiving data or responses in all the waiting states of a gateway, shows the best performance in all of data transmission real-timeliness, data loss and energy consumption.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.150-152
• /
• 2022

In this work, we propose a model that considers the behavior and synaptic plasticity of sensory neurons based on Liquid Time-constant Network (LTC). The neuron connection structure was experimented with four types: the increasing number of neurons, the decreasing number, the decreasing number, and the decreasing number. In this study, we experimented using a time series prediction dataset to see if the performance of the changed model improved compared to LTC. Experimental results show that the application of modeling of sensory neurons does not always bring about performance improvements, but improves performance through proper selection of learning rules depending on the type of dataset. In addition, the connective structure of neurons showed improved performance when it was less than four layers.