• Journal of Korean Society for Geospatial Information Science
• /
• v.18 no.2
• /
• pp.29-33
• /
• 2010

The purpose of LiDAR ground classification is to archive both goals which are acquiring confident ground points with high precision and describing ground shape in detail. In spite of many studies about developing optimized algorithms to kick out this, it is very difficult to classify ground points and describing ground shape by airborne LiDAR data. Especially it is more difficult in a dense forested area like Korea. Principle misclassification was mainly caused by complex forest canopy hierarchy in Korea and relatively coarse LiDAR points density for ground classification. Unfortunately, a lot of LiDAR surveying performed in summer in South Korea. And by that reason, schematic LiDAR points distribution is very different from those of Europe. So, this study propose enhanced ground classification method considering Korean land cover characteristics. Firstly, this study designate highly confident candidated LiDAR points as a first ground points which is acquired by using big roller classification algorithm. Secondly, this study applied weighted gradient kernel(WGK) algorithm to find and include highly expected ground points from the remained candidate points. This study methods is very useful for reconstruct deformed terrain due to misclassification results by detecting and include important terrain model key points for describing ground shape at site. Especially in the case of deformed bank side of river area, this study showed highly enhanced classification and reconstruction results by using WGK algorithm.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.11
• /
• pp.147-155
• /
• 2022

In this study, we developed a system to dynamically balance a daily stock portfolio and performed trading simulations using gradient boosting and genetic algorithms. We collected various stock market data from stocks listed on the KOSPI and KOSDAQ markets, including investor-specific transaction data. Subsequently, we indexed the data as a preprocessing step, and used feature engineering to modify and generate variables for training. First, we experimentally compared the performance of three popular gradient boosting algorithms in terms of accuracy, precision, recall, and F1-score, including XGBoost, LightGBM, and CatBoost. Based on the results, in a second experiment, we used a LightGBM model trained on the collected data along with genetic algorithms to predict and select stocks with a high daily probability of profit. We also conducted simulations of trading during the period of the testing data to analyze the performance of the proposed approach compared with the KOSPI and KOSDAQ indices in terms of the CAGR (Compound Annual Growth Rate), MDD (Maximum Draw Down), Sharpe ratio, and volatility. The results showed that the proposed strategies outperformed those employed by the Korean stock market in terms of all performance metrics. Moreover, our proposed LightGBM model with a genetic algorithm exhibited competitive performance in predicting stock price movements.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2010.05a
• /
• pp.377-378
• /
• 2010

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10a
• /
• pp.1017-1021
• /
• 1991

This paper is to explore the potential use of the modeling capacity of neural networks for control applications. The tasks are carried out by two neural networks which act as a plant identifier and a system controller, respectively. Using information stored in the identification network control action has been developed. Without supervising control signals are generated by a gradient type iterative algorithm.

• 제어로봇시스템학회:학술대회논문집
• /
• 1989.10a
• /
• pp.1097-1112
• /
• 1989

An iterative learning control scheme is incorporated to the computed torque method as a means to enhance the accuracy and the flexibility. A learning rule is constructed by utilizing a gradient descent algorithm and data compressing techniques are illustrated. Computer simulation results show a good performance of the scheme under a relatively high speed and a heavy payload condition.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.8
• /
• pp.126-135
• /
• 2014

The Euclidean distance between error probability density functions (EDEP) has been used as a performance criterion for supervised adaptive signal processing in impulsive noise environments. One of the drawbacks of the EDEP algorithm is a heavy computational complexity due to the double summation operations at each iteration time. In this paper, a recursive method to reduce its computational burden in the estimation of the EDEP and its gradient is proposed. For the data block size N, the computational complexity for the estimation of the EDEP and its gradient can be reduced to O(N) by the proposed method, while the conventional estimation method has $O(N^2)$. In the performance test, the proposed EDEP and its gradient estimation yield the same estimation results in the steady state as the conventional block-processing method. The simulation results indicates that the proposed method can be effective in practical adaptive signal processing.

• Journal of Broadcast Engineering
• /
• v.25 no.2
• /
• pp.157-165
• /
• 2020

In this paper. we propose an effective compensation scheme to the quantization error arisen from quantized learning in a machine learning on an embedded system. In the machine learning based on a gradient descent or nonlinear signal processing, the quantization error generates early vanishing of a gradient and occurs the degradation of learning performance. To compensate such quantization error, we derive an orthogonal compensation vector with respect to a maximum component of the gradient vector. Moreover, instead of the conventional constant learning rate, we propose the adaptive learning rate algorithm without any inner loop to select the step size, based on a nonlinear optimization technique. The simulation results show that the optimization solver based on the proposed quantized method represents sufficient learning performance.

• The Journal of the Acoustical Society of Korea
• /
• v.32 no.3
• /
• pp.279-285
• /
• 2013

In this paper, we propose a novel speech enhancement method to improve the performance of the conventional global soft decision which is based on the spectral gradient method applied to the ratio of a priori speech absence and presence probability value (q). Conventional global soft decision scheme used a fixed value of q in accordance with the hypothesis assumed, but the proposed algorithm is a technique for improving the speech absence probability which is applied adaptively variable value of q according to the speech presence or absence in the previous two frames and the conditions of the spectral gradient value. Experimental results show that the proposed improved global soft decision method based on the spectral gradient method yields better results compared to the conventional global soft decision technique based on the performance criteria of the ITU-T P. 862 PESQ (Perceptual Evaluation of Speech Quality).

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.10 no.2
• /
• pp.192-197
• /
• 2017

In this paper, image filter optimization method based on common sub-expression elimination is proposed for low-power image feature extraction hardware design. Low power and high performance object recognition hardware is essential for industrial robot which is used for factory automation. However, low area Gaussian gradient filter hardware design is required for object recognition hardware. For the hardware complexity reduction, we adopt the symmetric characteristic of the filter coefficients using the transposed form FIR filter hardware architecture. The proposed hardware architecture can be implemented without degradation of the edge detection data quality since the proposed hardware is implemented with original Gaussian gradient filtering algorithm. The expremental result shows the 50% of multiplier savings compared with previous work.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.22 no.10
• /
• pp.2317-2328
• /
• 1997

The objective of this paper is to develop an algorithm for a real-time automatic fingerprint recognition system. The algorithm employs the wavelet transform(WT) and the dominat local orientation that derived from the gradient Gaussian(GoG) and coherence in determining the directions of ridges in fingerprint images. By using the WT, the algorithm does not require conventional preprocessing procedures such as smothing, binarization, thining and restoration. For recognition, two fingerprint images are compared in three different ST domains;one that represents the original image compressed to quarter(LL), another that shows vertical directional characteristic(LH), and third as the block that contains horizontal direction(HL) in WT domain. Each block has dominat local orientation that derived from the GoG and coherence. The proposed algorithm is imprlemented on a SunSparc-2 workstation under X-window environment. Our simulation results, in real-time have shown that while the rate of Type II error-Incorrect recognition of two identical fingerprints as the identical fingerprints-is held at 0%, the rate of Type I error-Incorrect recognitionof two identical fingerprints as the different ones-is 2.5%.