• The Bulletin of The Korean Astronomical Society
• /
• v.40 no.1
• /
• pp.56.2-56.2
• /
• 2015

The KMTNet Supernovae Project utilizes the large $2^{\circ}{\times}2^{\circ}$ field of view of the three KMTNet telescopes to search and monitor supernovae, especially early ones, and other optical transients. A key component of the project is to build a data pipeline with a descent latency and an early alerting system that can handle the large volume of the data in an efficient and a prompt way, while minimizing false alarms, which casts a significant challenge to the software development. Here we present the current status of their development. The pipeline utilizes a difference image analysis technique to discover candidate transient sources after making correction of image distortion. In the early phase of the program, final selection of transient sources from candidates will mainly rely on multi-filter, multi-epoch and multi-site screening as well as human inspection, and an interactive web-based system is being developed for this purpose. Eventually, machine learning algorithms, based on the training set collected in the early phase, will be used to select true transient sources from candidates.

• Journal of Electrical Engineering and Technology
• /
• v.10 no.1
• /
• pp.408-421
• /
• 2015

Because the wheel of V-belt continuously variable transmission (CVT) system driven by permanent magnet synchronous motor (PMSM) has much unknown nonlinear and time-varying characteristics, the better control performance design for the linear control design is a time consuming job. In order to overcome difficulties for design of the linear controllers, a hybrid recurrent Chebyshev neural network (NN) control system is proposed to control for a PMSM servo-driven V-belt CVT system under the occurrence of the lumped nonlinear load disturbances. The hybrid recurrent Chebyshev NN control system consists of an inspector control, a recurrent Chebyshev NN control with adaptive law and a recouped control. Moreover, the online parameters tuning methodology of adaptive law in the recurrent Chebyshev NN can be derived according to the Lyapunov stability theorem and the gradient descent method. Furthermore, the optimal learning rate of the parameters based on discrete-type Lyapunov function is derived to achieve fast convergence. The recurrent Chebyshev NN with fast convergence has the online learning ability to respond to the system's nonlinear and time-varying behaviors. Finally, to show the effectiveness of the proposed control scheme, comparative studies are demonstrated by experimental results.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.546-551
• /
• 2004

In this paper, a level controller is designed with the neuro-fuzzy model based on Takagi-Sugeno fuzzy system. The fuzzy system is employed as the controller, which can be tuned by the neural network mechanism based on a gradient descent technique. The tuning mechanism will provide an optimal process input by forcing the process error to zero. The proposed controller provides the online tunable mode to adjust the consequent membership function parameters. The controller is implemented with M-file and graphic user interface (GUI) of Matlab program. The program uses MPIBM3 interface card to connect with the industrial processes In the experimentation, the proposed method is tested to vary of the process parameters, set points and load disturbance. Processes of one tank and two tanks are used to evaluate the efficiency of our controller. The results of the both processes are compared with two PID systems that are 3G25A-PIDO1-E and E5AK of OMRON. From the comparison results, our controller performance can be archived in the case of more robustness than the two PID systems.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.635-639
• /
• 2004

In this paper, we present a novel approach for the structure of Fuzzy Neural Network(FNN) based on wavelet function and apply this network structure to the modeling of chaotic nonlinear systems. Generally, the wavelet fuzzy model(WFM) has the advantage of the wavelet transform by constituting the fuzzy basis function(FBF) and the conclusion part to equalize the linear combination of FBF with the linear combination of wavelet functions. However, it is very difficult to identify the fuzzy rules and to tune the membership functions of the fuzzy reasoning mechanism. Neural networks, on the other hand, utilize their learning capability for automatic identification and tuning. Therefore, we design a wavelet based FNN structure(WFNN) that merges these advantages of neural network, fuzzy model and wavelet transform. The basic idea of our wavelet based FNN is to realize the process of fuzzy reasoning of wavelet fuzzy system by the structure of a neural network and to make the parameters of fuzzy reasoning be expressed by the connection weights of a neural network. And our network can automatically identify the fuzzy rules by modifying the connection weights of the networks via the gradient descent scheme. To verify the efficiency of our network structure, we evaluate the modeling performance for chaotic nonlinear systems and compare it with those of the FNN and the WFM.

• International Journal of Control, Automation, and Systems
• /
• v.6 no.5
• /
• pp.755-766
• /
• 2008

This study presents a recurrent compensatory fuzzy neural network (RCFNN) for dynamic system identification. The proposed RCFNN uses a compensatory fuzzy reasoning method, and has feedback connections added to the rule layer of the RCFNN. The compensatory fuzzy reasoning method can make the fuzzy logic system more effective, and the additional feedback connections can solve temporal problems as well. Moreover, an online learning algorithm is demonstrated to automatically construct the RCFNN. The RCFNN initially contains no rules. The rules are created and adapted as online learning proceeds via simultaneous structure and parameter learning. Structure learning is based on the measure of degree and parameter learning is based on the gradient descent algorithm. The simulation results from identifying dynamic systems demonstrate that the convergence speed of the proposed method exceeds that of conventional methods. Moreover, the number of adjustable parameters of the proposed method is less than the other recurrent methods.

• Proceedings of the IEEK Conference
• /
• 2002.07a
• /
• pp.30-33
• /
• 2002

To obtain good performance of radial basis function (RBF) neural networks, it needs very careful consideration in design. The selection of several parameters such as the number of centers and widths of the radial basis functions must be considered carefully since they critically affect the network's performance. We propose a learning algorithm for growing of complexity of RBF neural networks which is adapted automatically according to the complexity of tasks. The algorithm generates a new basis function based on the errors of network, the percentage of decreasing rate of errors and the nearest distance from input data to the center of hidden unit. The RBF's center is located at the point where the maximum of absolute interference error occurs in the input space. The width is calculated based on the standard deviation of distance between the center and inputs data. The steepest descent method is also applied for adjusting the weights, centers, and widths. To demonstrate the performance of the proposed algorithm, general problem of function estimation is evaluated. The results obtained from the simulation show that the proposed algorithm for RBF neural networks yields good performance in terms of convergence and accuracy compared with those obtained by conventional multilayer feedforward networks.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.7
• /
• pp.170-178
• /
• 1994

The nearest neighbor classification rule is widely used because it is not only simple but the error rate is asymptotically less than twice Bayes theoretical minimum error. But the method basically use the whole training patterns as the reference vectors. so that both storage and classification time increase as the number of training patterns increases. LVQ(Learning Vector Quantization) resolved this problem by training the reference vectors instead of just storing the whole training patterns. But it is a heuristic algorithm which has no theoretic background there is no terminating condition and it requires a lot of iterations to get to meaningful result. This paper is to propose a new training method of the reference vectors. which minimize the given error function. The nearest neighbor network,the network version of the nearest neighbor classification rule is proposed. The network is funtionally identical to the nearest neighbor classification rule is proposed. The network is funtionally identical to the nearest neighbor classification rule and the reference vectors are represented by the weights between the nodes. The network is trained to minimize the error function with respect to the weights by the steepest descent method. The learning algorithm is derived and it is shown that the proposed method can adjust more reference vectors than LVQ in each iteration. Experiment showed that the proposed method requires less iterations and the error rate is smaller than that of LVQ2.

• Korean Journal of Applied Biomechanics
• /
• v.27 no.3
• /
• pp.181-187
• /
• 2017

Objective: The aim of the study was to compare & analyze on the variations of ground reaction force during ascending and descending of bus stair. Method: Simulated wooden stair of bus (raiser: 37.66 cm, width: 109 cm, tread: 29 cm) and GRF system (AMTI-OR-7/ AMTI., USA) was set up within experimental room. Adult female (n=8) performed ascending & descending of simulated bus stair, and variables analyzed consisted of TT (transfer-time), PVF (peak vertical force), LR (loading rate), DR (decay rate), CV (coefficient of variation) and AI (asymmetry index). Sample data from GRF cut off at 1,000 Hz. Results: TT showed shortest variation at phase 1 during descending, but longest variation at phase 1 during ascending of stair. PVF19 (Fz2, 100%) showed large pattern during descending than that of ascending, but rather showed small pattern during ascending of stair in case of PVF2 (Fz4). LR showed larger pattern during descending than that of ascending, but rather during ascending of stair in case of DR. Variation of CV (%) did not show difference between LR and DR, but showed higher possible occurrence of variation during descending of stair. Also AI (%) showed higher index during ascending than that of descending of stair. Conclusion: Because introduction of lowered bus stair has various realistic problems, if lined up at designated bus stopage exactly, rather can solve problems of inconvenience, reduce impulsive force and secure a stability of COG during ascending & descending of stair.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.6
• /
• pp.3029-3045
• /
• 2017

The main processes of a cognitive radio system include spectrum sensing, spectrum decision, spectrum sharing, and spectrum conversion. Experimental results show that these stages introduce a time delay that affects the spectrum sensing accuracy, reducing its efficiency. To reduce the time delay, the frequency spectrum prediction was proposed to alleviate the burden on the spectrum sensing. In this paper, the deep recurrent neural network (DRNN) was proposed to predict the spectrum of multiple time slots, since the existing methods only predict the spectrum of one time slot. The continuous state of a channel is divided into a many time slots, forming a time series of the channel state. Since there are more hidden layers in the DRNN than in the RNN, the DRNN has fading memory in its bottom layer as well as in the past input. In addition, the extended Kalman filter was used to train the DRNN, which overcomes the problem of slow convergence and the vanishing gradient of the gradient descent method. The spectrum prediction based on the DRNN was verified with a WiFi signal, and the error of the prediction was analyzed. The simulation results proved that the multiple slot spectrum prediction improved the spectrum efficiency and reduced the energy consumption of spectrum sensing.

• Journal of the Korean Society of Visualization
• /
• v.15 no.2
• /
• pp.33-40
• /
• 2017

Abstract In the present study, the characteristics of kinematics and dynamics in the fluttering motion of free-falling plates are investigated at Reynolds number of $10^5$. We record quasi-two-dimensional trajectories of free-falling plates with and without superhydrophobic coating using high-speed camera, and compute the drag and lift forces by trajectory analysis. Translational and angular velocities are modeled as harmonic functions with specific phase differences. In particular, periodic mass elevations near turning points are explained using the suggested models. At each turning point, a sudden drop in lift and a rapid increase in drag occur simultaneously due to fast increase in angle of attack. However, the lift is increased over the buoyancy-corrected weight of plate during gliding flight, resulting in periodic mass elevations near turning points. Superhydrophobicity is shown to increase lift but to reduce drag on a fluttering plate, resulting in the decrease of mean descent speed.