• Proceedings of the Korea Water Resources Association Conference
• /
• 2020.06a
• /
• pp.170-170
• /
• 2020

Due to the development of technologies, complex computation of huge data set is possible with a prevalent personal computer. Therefore, machine learning methods have been widely applied in the hydrologic field such as regression-based regional frequency analysis (RFA). The main purpose of this study is to compare two frameworks of RFA based on the artificial neural network (ANN) models: quantile regression technique (QRT-ANN) and parameter regression technique (PRT-ANN). As an output layer of the ANN model, the QRT-ANN predicts quantiles for various return periods whereas the PRT-ANN provides prediction of three parameters for the generalized extreme value distribution. Rainfall gauging sites where record length is more than 20 years were selected and their annual maximum rainfalls and various hydro-meteorological variables were used as an input layer of the ANN model. While employing the ANN model, 70% and 30% of gauging sites were used as training set and testing set, respectively. For each technique, ANN model structure such as number of hidden layers and nodes was determined by a leave-one-out validation with calculating root mean square error (RMSE). To assess the performances of two frameworks, RMSEs of quantile predicted by the QRT-ANN are compared to those of the PRT-ANN.

• Steel and Composite Structures
• /
• v.47 no.6
• /
• pp.759-779
• /
• 2023

The paper proposes two hybrid metaheuristic optimization and artificial neural network (ANN) methods for the close prediction of the ultimate axial compressive capacity of concentrically loaded concrete filled double skin steel tube (CFDST) columns. Two metaheuristic optimization, namely genetic algorithm (GA) and particle swarm optimization (PSO), approaches enable the dynamic training architecture underlying an ANN model by optimizing the number and sizes of hidden layers as well as the weights and biases of the neurons, simultaneously. The former is termed as GA-ANN, and the latter as PSO-ANN. These techniques utilize the gradient-based optimization with Bayesian regularization that enhances the optimization process. The proposed GA-ANN and PSO-ANN methods construct the predictive ANNs from 125 available experimental datasets and present the superior performance over standard ANNs. Both the hybrid GA-ANN and PSO-ANN methods are encoded within a user-friendly graphical interface that can reliably map out the accurate ultimate axial compressive capacity of CFDST columns with various geometry and material parameters.

• Steel and Composite Structures
• /
• v.44 no.6
• /
• pp.867-882
• /
• 2022

Predicting the compressive strength of concrete (CSoC) is of high significance in civil engineering. The CSoC is a highly dependent and non-linear parameter that requires powerful models for its simulation. In this work, two novel optimization techniques, namely evaporation rate-based water cycle algorithm (ER-WCA) and equilibrium optimizer (EO) are employed for optimally finding the parameters of a multi-layer perceptron (MLP) neural processor. The efficiency of these techniques is examined by comparing the results of the ensembles to a conventionally trained MLP. It was observed that the ER-WCA and EO optimizers can enhance the training accuracy of the MLP by 11.18 and 3.12% (in terms of reducing the root mean square error), respectively. Also, the correlation of the testing results climbed from 78.80% to 82.59 and 80.71%. From there, it can be deduced that both ER-WCA-MLP and EO-MLP can be promising alternatives to the traditional approaches. Moreover, although the ER-WCA enjoys a larger accuracy, the EO was more efficient in terms of complexity, and consequently, time-effectiveness.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.9
• /
• pp.2483-2504
• /
• 2023

Most existing low-light enhancement algorithms either use a large number of training parameters or lack generalization to real-world scenarios. This paper presents a novel lightweight and robust pixel-wise polynomial approximation-based deep network for low-light image enhancement. For mapping the low-light image to the enhanced image, pixel-wise higher-order polynomials are employed. A deep convolution network is used to estimate the coefficients of these higher-order polynomials. The proposed network uses multiple branches to estimate pixel values based on different receptive fields. With a smaller receptive field, the first branch enhanced local features, the second and third branches focused on medium-level features, and the last branch enhanced global features. The low-light image is downsampled by the factor of 2b-1 (b is the branch number) and fed as input to each branch. After combining the outputs of each branch, the final enhanced image is obtained. A comprehensive evaluation of our proposed network on six publicly available no-reference test datasets shows that it outperforms state-of-the-art methods on both quantitative and qualitative measures.

• Korean Journal of Artificial Intelligence
• /
• v.11 no.4
• /
• pp.1-8
• /
• 2023

In this paper, we propose a black ice detection platform framework using Convolutional Neural Networks (CNNs). To overcome black ice problem, we introduce a real-time based early warning platform using CNN-based architecture, and furthermore, in order to enhance the accuracy of black ice detection, we apply a multi-scale dilation convolution feature fusion (MsDC-FF) technique. Then, we establish a specialized experimental platform by using a comprehensive dataset of thermal road black ice images for a training and evaluation purpose. Experimental results of a real-time black ice detection platform show the better performance of our proposed network model compared to conventional image segmentation models. Our proposed platform have achieved real-time segmentation of road black ice areas by deploying a road black ice area segmentation network on the edge device Jetson Nano devices. This approach in parallel using multi-scale dilated convolutions with different dilation rates had faster segmentation speeds due to its smaller model parameters. The proposed MsCD-FF Net(2) model had the fastest segmentation speed at 5.53 frame per second (FPS). Thereby encouraging safe driving for motorists and providing decision support for road surface management in the road traffic monitoring department.

• Journal of information and communication convergence engineering
• /
• v.21 no.4
• /
• pp.300-305
• /
• 2023

Audience reactions in response to remote virtual performances must be compressed before being transmitted to the server. The server, which aggregates these data for group insights, requires a distribution code for the transfer. Recently, distributed learning algorithms such as federated learning have gained attention as alternatives that satisfy both the information security and efficiency requirements. In distributed learning, no individual user has access to complete information, and the objective is to achieve a learning effect similar to that achieved with the entire information. It is therefore important to distribute interdependent information among users and subsequently aggregate this information following training. In this paper, we present a new extension technique for minimal code that allows a new minimal code with a different length and Hamming weight to be generated through the product of any vector and a given minimal code. Thus, the proposed technique can generate minimal codes with previously unknown parameters. We also present a scenario wherein these combined methods can be applied.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.6
• /
• pp.319-325
• /
• 2023

In this study, we proposed a method for hyperparameter optimization in the building and training of a deep learning model designed to process point cloud data collected by a millimeter-wave radar system. The primary aim of this study is to facilitate the deployment of a baseline model in resource-constrained IoT devices. We evaluated a RadHAR baseline deep learning model trained on a public dataset composed of point clouds representing five distinct human activities. Additionally, we introduced a coarse-to-fine hyperparameter optimization procedure, showing substantial potential to enhance model efficiency without compromising predictive performance. Experimental results show the feasibility of significantly reducing model size without adversely impacting performance. Specifically, the optimized model demonstrated a 3.3% improvement in classification accuracy despite a 16.8% reduction in number of parameters compared th the baseline model. In conclusion, this research offers valuable insights for the development of deep learning models for resource-constrained IoT devices, underscoring the potential of hyperparameter optimization and model size reduction strategies. This work contributes to enhancing the practicality and usability of deep learning models in real-world environments, where high levels of accuracy and efficiency in data processing and classification tasks are required.

• The Journal of Korea Robotics Society
• /
• v.19 no.2
• /
• pp.203-212
• /
• 2024

Simultaneous motion is essential in the activities of daily living (ADL). For motion intention recognition, surface electromyogram (sEMG) and corresponding motion label is necessary. However, this process is time-consuming and it may increase the burden of the user. Therefore, we propose a simultaneous motion recognition framework using data augmentation based on muscle activation model. The model consists of multiple point sources to be optimized while the number of point sources and their initial parameters are automatically determined. From the experimental results, it is shown that the framework has generated the data which are similar to the real one. This aspect is quantified with the following two metrics: structural similarity index measure (SSIM) and mean squared error (MSE). Furthermore, with k-nearest neighbor (k-NN) or support vector machine (SVM), the classification accuracy is also enhanced with the proposed framework. From these results, it can be concluded that the generalization property of the training data is enhanced and the classification accuracy is increased accordingly. We expect that this framework reduces the burden of the user from the excessive and time-consuming data acquisition.

• Computers and Concrete
• /
• v.33 no.2
• /
• pp.137-145
• /
• 2024

The technique of experimentally determining concrete's compressive strength for a given mix design is time-consuming and difficult. The goal of the current work is to propose a best working predictive model based on different machine learning algorithms such as Gradient Boosting Machine (GBM), Stacked Ensemble (SE), Distributed Random Forest (DRF), Extremely Randomized Trees (XRT), Generalized Linear Model (GLM), and Deep Learning (DL) that can forecast the compressive strength of ternary geopolymer concrete mix without carrying out any experimental procedure. A geopolymer mix uses supplementary cementitious materials obtained as industrial by-products instead of cement. The input variables used for assessing the best machine learning algorithm not only include individual ingredient quantities, but molarity of the alkali activator and age of testing as well. Myriad statistical parameters used to measure the effectiveness of the models in forecasting the compressive strength of ternary geopolymer concrete mix, it has been found that GBM performs better than all other algorithms. A sensitivity analysis carried out towards the end of the study suggests that GBM model predicts results close to the experimental conditions with an accuracy between 95.6 % to 98.2 % for testing and training datasets.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.24 no.1
• /
• pp.181-187
• /
• 2024

There are two kinds of well-known CNN methods, the group equivariant CNN and the CNN using steerable filters, which have excellent classification performances for randomly rotated objects in image space. This paper describes their mathematical structures and introduces implementation methods. We implement them, including the existing CNN, which have the same number of filters, then compare and analyze their performances by simulating them with the randomly rotated MNIST. According to the experimental results, the steerable CNN, which shows a classification improvement over the others, has a relatively small number of parameters to learn, so performance degradation is relatively small even when the size of the training dataset is reduced.