• Wind and Structures
• /
• v.18 no.6
• /
• pp.619-631
• /
• 2014

This paper proposes new criteria to fix hidden neuron in Multilayer Perceptron Networks for wind speed prediction in renewable energy systems. To fix hidden neurons, 101 various criteria are examined based on the estimated mean squared error. The results show that proposed approach performs better in terms of testing mean squared errors. The convergence analysis is performed for the various proposed criteria. Mean squared error is used as an indicator for fixing neuron in hidden layer. The proposed criteria find solution to fix hidden neuron in neural networks. This approach is effective, accurate with minimal error than other approaches. The significance of increasing the number of hidden neurons in multilayer perceptron network is also analyzed using these criteria. To verify the effectiveness of the proposed method, simulations were conducted on real time wind data. Simulations infer that with minimum mean squared error the proposed approach can be used for wind speed prediction in renewable energy systems.

• Journal of KIISE
• /
• v.45 no.1
• /
• pp.69-75
• /
• 2018

In the streaming data where data samples arrive sequentially in time, it is difficult to apply the dimension reduction method based on batch learning. Therefore an incremental dimension reduction method for the application to streaming data has been studied. In this paper, we propose an incremental linear discriminant analysis method using the least squared error solution. Instead of computing scatter matrices directly, the proposed method incrementally updates the projective direction for dimension reduction by using the information of a new incoming sample. The experimental results demonstrate that the proposed method is more efficient compared with previously proposed incremental dimension reduction methods.

• ETRI Journal
• /
• v.37 no.3
• /
• pp.533-540
• /
• 2015

In wireless localization, several linear closed-form solution (LCS) methods have been investigated as a direct result of the drawbacks that plague the existing iterative methods, such as the local minimum problem and heavy computational burden. Among the known LCS methods, both the direct solution method and the difference of squared range measurements method are considered in this paper. These LCS methods do not have any of the aforementioned problems that occur in the existing iterative methods. However, each LCS method does have its own individual error property. In this paper, a hybrid LCS method is presented to reduce these errors. The hybrid LCS method integrates the two aforementioned LCS methods by using two check points that give important information on the probability of occurrence of each LCS's individual error. The results of several Monte Carlo simulations show that the proposed method has a good performance. The solutions provided by the proposed method are accurate and reliable. The solutions do not have serious errors such as those that occur in the conventional standalone LCS and iterative methods.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.8
• /
• pp.896-902
• /
• 2016

In this paper, we investigate a minimum mean-squared error based nonlinear successive precoding method as a practical solution of dirty paper coding for multiuser downlink channels where each user has more than one antenna in the presence of other cell interference (OCI). Unlike conventional zero-forcing (ZF) based methods, the proposed scheme takes the OCI plus noise into account when suppressing the inter-cell multiuser interference, which results in improvement of the received signal-to-interference-plus-noise ratio. Simulation results show that the proposed scheme outperforms conventional methods in terms of sum rate for various OCI configurations.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.10
• /
• pp.1059-1066
• /
• 2011

This paper makes a close inquiry into ill-conditioning that may be occurred in wireless localization of the sensor nodes based on network signals in the wireless sensor network and provides the clue for solving the problem. In order to estimate the location of a node based on the range information calculated using the signal propagation time, LS (Least Squares) method is usually used. The LS method estimates the solution that makes the squared estimation error minimal. When a nonlinear function is used for the wireless localization, ILS (Iterative Least Squares) method is used. The ILS method process the LS method iteratively after linearizing the nonlinear function at the initial nominal point. This method, however, has a problem that the final solution may converge into a LM (Local Minimum) instead of a GM (Global Minimum) according to the deployment of the fixed nodes and the initial nominal point. The conditions that cause the problem are explained and an adaptive method is presented to solve it, in this paper. It can be expected that the stable location solution can be provided in implementation of the wireless localization methods based on the results of this paper.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2021.06a
• /
• pp.373-373
• /
• 2021

Rainfall forecasting is an important issue that is applied in many areas, such as agriculture, flood warning, and water resources management. In this context, this study proposed a statistical and machine learning-based forecasting model for monthly rainfall. The Bayesian Gaussian process was chosen to optimize the hyperparameters of the Stacked Long Short-term memory (SLSTM) model. The proposed SLSTM model was applied for predicting monthly precipitation of Seoul station, South Korea. Data were retrieved from the Korea Meteorological Administration (KMA) in the period between 1960 and 2019. Four schemes were examined in this study: (i) prediction with only rainfall; (ii) with deseasonalized rainfall; (iii) with rainfall and minimum temperature; (iv) with deseasonalized rainfall and minimum temperature. The error of predicted rainfall based on the root mean squared error (RMSE), 16-17 mm, is relatively small compared with the average monthly rainfall at Seoul station is 117mm. The results showed scheme (iv) gives the best prediction result. Therefore, this approach is more straightforward than the hydrological and hydraulic models, which request much more input data. The result indicated that a deep learning network could be applied successfully in the hydrology field. Overall, the proposed method is promising, given a good solution for rainfall prediction.

• Journal of Institute of Control, Robotics and Systems
• /
• v.19 no.7
• /
• pp.633-639
• /
• 2013

Several estimation methods used in the range measurement based wireless localization area have individual problems. These problems may not occur according to certain application areas. However, these problems may give rise to serious problems in particular applications. In this paper, three methods, ILS (Iterative Least Squares), DS (Direct Solution), and DSRM (Difference of Squared Range Measurements) methods are considered. Problems that can occur in these methods are defined and a simple hybrid solution is presented to solve them. The ILS method is the most frequently used method in wireless localization and has local minimum problems and a large computational burden compared with closed-form solutions. The DS method requires less processing time than the ILS method. However, a solution for this method may include a complex number caused by the relations between the location of reference nodes and range measurement errors. In the near-field region of the complex solution, large estimation errors occur. In the DSRM method, large measurement errors occur when the mobile node is far from the reference nodes due to the combination of range measurement error and range data. This creates the problem of large localization errors. In this paper, these problems are defined and a hybrid localization method is presented to avoid them by integrating the DS and DSRM methods. The defined problems are confirmed and the performance of the presented method is verified by a Monte-Carlo simulation.

• Journal of the Optical Society of Korea
• /
• v.12 no.4
• /
• pp.249-254
• /
• 2008

In this paper we present an electronic domain solution for chromatic dispersion (CD) monitoring algorithm based on the estimated time domain channel in electronic domain using channel estimation methods. The proposed scheme utilizes kurtosis as a CD measurement, directly computed from the estimated inter-symbol-interference (ISI) channel due to the CD distortion. Hence, the proposed scheme exhibits robust performance under OSNR variation, in contrast to the existing electronic domain approach based on minimum mean squared error (MMSE) fractionally-spaced equalizer taps [1]. The simulation results verify the CD monitoring ability of the proposed scheme.

• Journal of Korea Water Resources Association
• /
• v.55 no.11
• /
• pp.903-911
• /
• 2022

Groundwater, one of the resources for supplying water, fluctuates in water level due to various natural factors. Recently, research has been conducted to predict fluctuations in groundwater levels using Artificial Neural Network (ANN). Previously, among operators in ANN, Gradient Descent (GD)-based Optimizers were used as Optimizer that affect learning. GD-based Optimizers have disadvantages of initial correlation dependence and absence of solution comparison and storage structure. This study developed Gradient Descent combined with Harmony Search (GDHS), a new Optimizer that combined GD and Harmony Search (HS) to improve the shortcomings of GD-based Optimizers. To evaluate the performance of GDHS, groundwater level at Icheon Yullhyeon observation station were learned and predicted using Multi Layer Perceptron (MLP). Mean Squared Error (MSE) and Mean Absolute Error (MAE) were used to compare the performance of MLP using GD and GDHS. Comparing the learning results, GDHS had lower maximum, minimum, average and Standard Deviation (SD) of MSE than GD. Comparing the prediction results, GDHS was evaluated to have a lower error in all of the evaluation index than GD.

• Journal of Advanced Navigation Technology
• /
• v.10 no.4
• /
• pp.371-376
• /
• 2006

The initial radius of sphere decoder has great effect on the bit error rate performance and computational complexity. Until now, it has been determined either by considering the statistical property of channel or by using of MMSE solution. The initial radius obtained by using statistical property of channel includes the lattice point corresponding to the transmit signal vector with very high probability. The method using MMSE solution first calculates out the MMSE solution of the received signal, then maps the hard decision of this solution into the received signal space, and finally the distance between the mapped point and the received signal is selected as the initial radius of the sphere decoding. In this paper, we derive a simple equation for initial radius selection which uses statistical property of channel and compare it with the method using MMSE solution. To compare two methods we define new metric 'Tightness'. Through the simulation, we observe that in low and moderate SNR region, the method using MMSE solution provides more complexity reduction for decoding while in high SNR region, the method using channel statistics is better.