• KIPS Transactions on Software and Data Engineering
• /
• v.13 no.1
• /
• pp.17-27
• /
• 2024

Video frame interpolation is an important technique used in the field of video and media, as it increases the continuity of motion and enables smooth playback of videos. In the study of video frame interpolation using deep learning, Kernel Based Method captures local changes well, but has limitations in handling global changes. In this paper, we propose a new U-Net structure that applies feature map differentiation and two directions to focus on capturing major changes to generate intermediate frames more accurately while reducing the number of parameters. Experimental results show that the proposed structure outperforms the existing model by up to 0.3 in PSNR with about 61% fewer parameters on common datasets such as Vimeo, Middle-burry, and a new YouTube dataset. Code is available at https://github.com/Go-MinSeong/SF-AdaCoF.

• Structural Engineering and Mechanics
• /
• v.63 no.1
• /
• pp.115-124
• /
• 2017

In this paper, a different technique to predict the effects of soil-structure interaction (SSI) on seismic response of building systems is investigated. The technique use a machine learning algorithm called Support Vector Regression (SVR) with technical and analytical results as input features. Normally, the effects of SSI on seismic response of existing building systems can be identified by different types of large data sets. Therefore, predicting and estimating the seismic response of building is a difficult task. It is possible to approximate a real valued function of the seismic response and make accurate investing choices regarding the design of building system and reduce the risk involved, by giving the right experimental and/or numerical data to a machine learning regression, such as SVR. The seismic response of both single-degree-of-freedom system and six-storey RC frame which can be represent of a broad range of existing structures, is estimated using proposed SVR model, while allowing flexibility of the soil-foundation system and SSI effects. The seismic response of both single-degree-of-freedom system and six-storey RC frame which can be represent of a broad range of existing structures, is estimated using proposed SVR model, while allowing flexibility of the soil-foundation system and SSI effects. The results show that the performance of the technique can be predicted by reducing the number of real data input features. Further, performance enhancement was achieved by optimizing the RBF kernel and SVR parameters through grid search.

• Journal of Information Processing Systems
• /
• v.12 no.3
• /
• pp.436-454
• /
• 2016

This research presents the battery discharge rate models for the energy consumption of mobile phone batteries based on machine learning by taking into account three usage patterns of the phone: the standby state, video playing, and web browsing. We present the experimental design methodology for collecting data, preprocessing, model construction, and parameter selections. The data is collected based on the HTC One X hardware platform. We considered various setting factors, such as Bluetooth, brightness, 3G, GPS, Wi-Fi, and Sync. The battery levels for each possible state vector were measured, and then we constructed the battery prediction model using different regression functions based on the collected data. The accuracy of the constructed models using the multi-layer perceptron (MLP) and the support vector machine (SVM) were compared using varying kernel functions. Various parameters for MLP and SVM were considered. The measurement of prediction efficiency was done by the mean absolute error (MAE) and the root mean squared error (RMSE). The experiments showed that the MLP with linear regression performs well overall, while the SVM with the polynomial kernel function based on the linear regression gives a low MAE and RMSE. As a result, we were able to demonstrate how to apply the derived model to predict the remaining battery charge.

• Journal of Information Processing Systems
• /
• v.14 no.1
• /
• pp.176-190
• /
• 2018

This paper presents a novel convolutional neural network based multi-feature fusion learning method for non-rigid 3D model retrieval, which can investigate the useful discriminative information of the heat kernel signature (HKS) descriptor and the wave kernel signature (WKS) descriptor. At first, we compute the 2D shape distributions of the two kinds of descriptors to represent the 3D model and use them as the input to the networks. Then we construct two convolutional neural networks for the HKS distribution and the WKS distribution separately, and use the multi-feature fusion layer to connect them. The fusion layer not only can exploit more discriminative characteristics of the two descriptors, but also can complement the correlated information between the two kinds of descriptors. Furthermore, to further improve the performance of the description ability, the cross-connected layer is built to combine the low-level features with high-level features. Extensive experiments have validated the effectiveness of the designed multi-feature fusion learning method.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.10b
• /
• pp.22-27
• /
• 2007

Phosphorylation is one of the most important post translational modifications which regulate the activity of proteins. The problem of predicting phosphorylation sites is the first step of understanding various biological processes that initiate the actual function of proteins in each signaling pathway. Although many prediction methods using single or multiple features extracted from protein sequences have been proposed, systematic data integration approach has not been applied in order to improve the accuracy of predicting general phosphorylation sites. In this paper, we propose an optimal way of integrating multiple features in the framework of multiple kernel learning. We optimally combine seven kernels extracted from sequence, physico-chemical properties, pairwise alignment, and structural information. Using the data set of Phospho. ELM, the accuracy evaluated by 5-fold cross-validation reaches 85% for serine, 85% for threonine, and 81% for tyrosine. Our computational experiments show significant improvement in the performance of prediction relative to a single feature, or to the combined feature with equal weights. Moreover, our systematic integration method significantly improves the prediction preformance compared with the previous well-known methods.

• Communications for Statistical Applications and Methods
• /
• v.31 no.2
• /
• pp.213-234
• /
• 2024

In recent decades, increasing research attention has been directed toward predicting the price of stocks in financial markets using deep learning methods. For instance, recurrent neural network (RNN) is known to be competitive for datasets with time-series data. Long short term memory (LSTM) further improves RNN by providing an alternative approach to the gradient loss problem. LSTM has its own advantage in predictive accuracy by retaining memory for a longer time. In this paper, we combine both supervised and unsupervised dimension reduction methods with LSTM to enhance the forecasting performance and refer to this as a dimension reduction based LSTM (DR-LSTM) approach. For a supervised dimension reduction method, we use methods such as sliced inverse regression (SIR), sparse SIR, and kernel SIR. Furthermore, principal component analysis (PCA), sparse PCA, and kernel PCA are used as unsupervised dimension reduction methods. Using datasets of real stock market index (S&P 500, STOXX Europe 600, and KOSPI), we present a comparative study on predictive accuracy between six DR-LSTM methods and time series modeling.

• Journal of KIISE
• /
• v.43 no.6
• /
• pp.636-644
• /
• 2016

Annotated Corpus is important to understand natural language using machine learning method. In this paper, we propose a new method to automate error reduction of annotated corpora. We use the Ripple-Down Rules(RDR) for reducing errors and Kernel to extend RDR for NLP. We applied our system to the Korean Wikipedia and blog corpus errors to find the annotated corpora error type. Experimental results with various views from the Korean Wikipedia and blog are reported to evaluate the effectiveness and efficiency of our proposed approach. The proposed approach can be used to reduce errors of large corpora.

• Korean Journal of Remote Sensing
• /
• v.36 no.4
• /
• pp.609-626
• /
• 2020

Satellite-based land surface temperature (LST) has been used as one of the major parameters in various climate and environmental models. Especially, Moderate Resolution Imaging Spectroradiometer (MODIS) LST is the most widely used satellite-based LST product due to its spatiotemporal coverage (1 km spatial and sub-daily temporal resolutions) and longevity (> 20 years). However, there is an increasing demand for LST products with finer spatial resolution (e.g., 10-250 m) over regions such as urban areas. Therefore, various methods have been proposed to produce high-resolution MODIS-like LST less than 250 m (e.g., 100 m). The purpose of this review is to provide a comprehensive overview of recent research trends and challenges for the downscaling of MODIS LST. Based on the recent literature survey for the past decade, the downscaling techniques classified into three groups-kernel-driven, fusion-based, and the combination of kernel-driven and fusion-based methods-were reviewed with their pros and cons. Then, five open issues and challenges were discussed: uncertainty in LST retrievals, low thermal contrast, the nonlinearity of LST temporal change, cloud contamination, and model generalization. Future research directions of LST downscaling were finally provided.

• Geomechanics and Engineering
• /
• v.38 no.1
• /
• pp.29-44
• /
• 2024

This research investigates the effectiveness of data augmentation techniques in the automated analysis of B-scan images from ground-penetrating radar (GPR) using deep learning. In spite of the growing interest in automating GPR data analysis and advancements in deep learning for image classification and object detection, many deep learning-based GPR data analysis studies have been limited by the availability of large, diverse GPR datasets. Data augmentation techniques are widely used in deep learning to improve model performance. In this study, we applied four data augmentation techniques (geometric transformation, color-space transformation, noise injection, and applying kernel filter) to the GPR datasets obtained from a testbed. A deep learning model for GPR data analysis was developed using three models (Faster R-CNN ResNet, SSD ResNet, and EfficientDet) based on transfer learning. It was found that data augmentation significantly enhances model performance across all cases, with the mAP and AR for the Faster R-CNN ResNet model increasing by approximately 4%, achieving a maximum mAP (Intersection over Union = 0.5:1.0) of 87.5% and maximum AR of 90.5%. These results highlight the importance of data augmentation in improving the robustness and accuracy of deep learning models for GPR B-scan analysis. The enhanced detection capabilities achieved through these techniques contribute to more reliable subsurface investigations in geotechnical engineering.

• ETRI Journal
• /
• v.31 no.5
• /
• pp.565-575
• /
• 2009

This paper concerns the use of support vector regression (SVR), which is based on the kernel method for learning from examples, in identification of walking robots. To handle complex dynamics in humanoid robot and realize stable walking, this paper develops and implements two types of reference natural motions for a humanoid, namely, walking trajectories on a flat floor and on an ascending slope. Next, SVR is applied to model stable walking motions by considering these actual motions. Three kinds of kernels, namely, linear, polynomial, and radial basis function (RBF), are considered, and the results from these kernels are compared and evaluated. The results show that the SVR approach works well, and SVR with the RBF kernel function provides the best performance. Plus, it can be effectively applied to model and control a practical biped walking robot.