• The Journal of the Acoustical Society of Korea
• /
• v.39 no.4
• /
• pp.246-254
• /
• 2020

In this paper, we propose a learning based compressive sensing algorithm for the purpose of side scan sonar image denoising. The proposed method is based on Iterative Shrinkage and Thresholding Algorithm (ISTA) framework and incorporates a powerful strategy that reinforces the non-linearity of deep learning network for improved performance. The proposed method consists of three essential modules. The first module consists of a non-linear transform for input and initialization while the second module contains the ISTA block that maps the input features to sparse space and performs inverse transform. The third module is to transform from non-linear feature space to pixel space. Superiority in noise removal and memory efficiency of the proposed method is verified through various experiments.

• Composites Research
• /
• v.14 no.5
• /
• pp.26-37
• /
• 2001

After impact analysis of the composite cylindrical shells was performed. obtained outputs at 9 equally divided points of the shell were used as input patterns of the neural networks. Identification of impact loading characteristics was predicted simultaneously. Momentum backpropagation algorithm of neural networks which can modify the momentum coefficient and learning rate was developed and applied to identify the loading characteristics. Hidden layers of the backpropagation increased from 1 layer to 3 layers and trained the loading characteristics. Developed program with variable learning rate was converged close to real load characteristics under 1% error. Inverse engineering which identify the impact loading characteristics can be applicable to the composite laminated cylindrical shells with developed neural networks.

• East Asian Economic Review
• /
• v.21 no.4
• /
• pp.385-432
• /
• 2017

This paper investigates how firm age, size and ownership are related with job creation and destruction, and how these patterns differ across transition and non-transition economies. The analysis finds that age is inversely related with gross job creation and net job creation in the two samples. This finding is consistent with the theory of the learning effect. The relationship between age and job destruction is indifferent in non-transition economies. On the contrary, old firms in transition economies destroy more jobs than young ones. The paper further establishes an inverse relationship between size and gross job creation in the two groups. However, there is divergence between the two samples; small firms in non-transition economies also exhibit a higher gross job destruction rate. Consequently large firms have a higher net job creation rate. In transition economies, small and large firms exhibit similar rates of job destruction. But small firms retain a higher net job creation rate. A more intriguing finding is that state owned firms do not underperform domestic private ones. This means these countries may be using soft budget constraint which allows state owned firms to overstaff. Finally, crowding out of SMEs by foreign owned firms is not evident in transition economies.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.18 no.5
• /
• pp.634-642
• /
• 1993

In this paper a new trajectory control method is proposed for a robot manipulator using a time delay neural network(TDNN) as a feedforward controller with an algorithm to learn inverse dynamics of the manipulator. The TDNN structure has so favorable characteristics that neurons can extract more dynamic information from both present and past input signals and perform more efficient learning. The TDNN neural network receives two normalized inputs, one of which is the reference trajectory signal and the other of which is the error signals from the PD controller. It is proved that the normalized inputs to the TDNN neural network can enhance the learning efficiency of the neural network. The proposed scheme was investigated for the planar robot manipulator with two joints by computer simulation.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.8
• /
• pp.186-192
• /
• 1999

This paper introduces a study of a method for the forward kinematic analysis, which finds the 6 DOF motions and velocities from the given six cylinder lengths in the Stewart platform. From the viewpoints of kinematics, the solution for the inverse kinematic is easily found by using the vectors of the links which are composed of the joint coordinates in base and plate frames, to act contrary to the serial manipulator, but forward kinematic is difficult because of the nonlinearity and complexity of the Stewart platform dynamic equation with the multi-solutions. Hence we, first in this study, introduce the linear estimator using the Luenberger's observer, and the estimator using the nonlinear measured model for the forward kinematic solutions. But it is difficult to find the parameter of the design for the estimation gain or to select the estimation gain and the constant steady state error exists. So this study suggests the estimator with the estimation gain to be learned by the neural network with the structure of multi-perceptron and the learning method using back propagation and shows the estimation performance using the simulation.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.8 no.3
• /
• pp.185-191
• /
• 2008

In this paper, experimental studies of a neural network (NN) control technique for non-model based position control of the x-y table robot are presented. Decentralized neural networks are used to control each axis of the x-y table robot separately. For an each neural network compensator, an inverse control technique is used. The neural network control technique called the reference compensation technique (RCT) is conceptually different from the existing neural controllers in that the NN controller compensates for uncertainties in the dynamical system by modifying desired trajectories. The back-propagation learning algorithm is developed in a real time DSP board for on-line learning. Practical real time position control experiments are conducted on the x-y table robot. Experimental results of using neural networks show more excellent position tracking than that of when PD controllers are used only.

• Journal of Information Processing Systems
• /
• v.17 no.4
• /
• pp.818-833
• /
• 2021

To address the problems of low precision rate, insufficient feature extraction, and poor contextual ability in existing text sentiment analysis methods, a mixed model account of a CNN-BiLSTM-TE (convolutional neural network, bidirectional long short-term memory, and topic extraction) model was proposed. First, Chinese text data was converted into vectors through the method of transfer learning by Word2Vec. Second, local features were extracted by the CNN model. Then, contextual information was extracted by the BiLSTM neural network and the emotional tendency was obtained using softmax. Finally, topics were extracted by the term frequency-inverse document frequency and K-means. Compared with the CNN, BiLSTM, and gate recurrent unit (GRU) models, the CNN-BiLSTM-TE model's F1-score was higher than other models by 0.0147, 0.006, and 0.0052, respectively. Then compared with CNN-LSTM, LSTM-CNN, and BiLSTM-CNN models, the F1-score was higher by 0.0071, 0.0038, and 0.0049, respectively. Experimental results showed that the CNN-BiLSTM-TE model can effectively improve various indicators in application. Lastly, performed scalability verification through a takeaway dataset, which has great value in practical applications.

• Journal of Korea Multimedia Society
• /
• v.22 no.1
• /
• pp.1-9
• /
• 2019

Single image rain removal is a typical inverse problem which decomposes the image into a background scene and a rain streak. Recent works have witnessed a substantial progress on the task due to the development of convolutional neural network (CNN). However, existing CNN-based approaches train the network with synthetically generated training examples. These data tend to make the network bias to the synthetic scenes. In this paper, we present an unsupervised framework for removing rain streaks from real-world rainy images. We focus on the natural phenomena that static rainy scenes capture a common background but different rain streak. From this observation, we train siamese network with the real rain image pairs, which outputs identical backgrounds from the pairs. To train our network, a real rainy dataset is constructed via web-crawling. We show that our unsupervised framework outperforms the recent CNN-based approaches, which are trained by supervised manner. Experimental results demonstrate that the effectiveness of our framework on both synthetic and real-world datasets, showing improved performance over previous approaches.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.1
• /
• pp.1-9
• /
• 2019

In this paper, we propose a method of generating a 2D gray image from LiDAR 3D reflection intensity. The proposed method uses the Fully Convolutional Network (FCN) to generate the gray image from 2D reflection intensity which is projected from LiDAR 3D intensity. Both encoder and decoder of FCN are configured with several convolution blocks in the symmetric fashion. Each convolution block consists of a convolution layer with $3{\times}3$ filter, batch normalization layer and activation function. The performance of the proposed method architecture is empirically evaluated by varying depths of convolution blocks. The well-known KITTI data set for various scenarios is used for training and performance evaluation. The simulation results show that the proposed method produces the improvements of 8.56 dB in peak signal-to-noise ratio and 0.33 in structural similarity index measure compared with conventional interpolation methods such as inverse distance weighted and nearest neighbor. The proposed method can be possibly used as an assistance tool in the night-time driving system for autonomous vehicles.

• 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.