• Journal of Korea Multimedia Society
• /
• v.6 no.1
• /
• pp.136-147
• /
• 2003

To achieve an optimal performance for a given problem, we need an integrative process of the parameter optimization via learning and the structure optimization via model selection. In this paper, we propose an efficient optimization method for improving generalization performance by considering the property of each sub-method and by combining them with common theoretical properties. First, weight parameters are optimized by natural gradient teaming with adaptive regularization, which uses a diverse error function. Second, the network structure is optimized by eliminating unnecessary parameters with natural pruning. Through iterating these processes, candidate models are constructed and evaluated based on the Bayesian Information Criterion so that an optimal one is finally selected. Through computational experiments on benchmark problems, we confirm the weight parameter and structure optimization performance of the proposed method.

• Korean Journal of Biological Psychiatry
• /
• v.6 no.1
• /
• pp.12-20
• /
• 1999

In order to understand the pathogenesis and progression of some synaptic loss related neuropsychiatric diseases, We attempted to develop a computer model in this study. We made a simple autoassociative memory network remembering numbers, transformed it into a disease model by pruning synapses, and measured its memory performance as a function of synaptic deletion. Decline in performance was measured as amount of synaptic loss increases and its mode of decline is sudden or gradual according to the mode of synaptic pruning. The developed computer model demonstrated how synaptic loss could cause memory impairment through a series of computer simulations, and suggested a new way of research in neuropsychiatry.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.44 no.5
• /
• pp.59-67
• /
• 2016

In landscape design by public institutions, although the costs and species of landscape trees stipulated by the Korean Public Procurement Service(PPS) are generally adhered to, the PPS regulations about planting trees with well-developed rootlets are almost entirely neglected. This study aimed to evaluate the performance of buried container modules, which are a new technology and product in landscape production that is able to reduce the defect rate while complying with regulations. To this end, this study measured rootlet density, rootlet development length, rootlet survival rate on excavation, and impairments of tree growth for 3 months after root pruning, and compared these variables for the container modules with those for trees that underwent root pruning in bare ground, and those that were cultivated in a container above ground. The results were as follows: First, the rootlet density was 88% for the trees in container modules, which was very high. Trees that underwent standard root pruning in bare ground had a somewhat lower density of 64%. Meanwhile, the trees that were cultivated in pots above ground died, invalidating measurement. Second, in terms of rootlet development and rootlet survival rate, the trees in container modules showed a mean length of 10.4cm, and 100% survival rate, indicating that there was no rootlet damage caused by excavation. For the trees that only underwent root pruning in bare ground, the mean length was 25.6cm and the rootlet survival rate was only half that of the trees in container modules, at 56%, demonstrating considerable damage. Rootlet development did not occur at all in the trees grown in pots. Third, the trees in container modules and those that underwent root pruning in bare ground did not show any deaths during the root pruning process, or any impairments such as stunted leaf growth. Conversely, the trees grown in pots nearly all died, and severe impairments of tree growth were observed. As shown by the results above, when we evaluated the performance of buried container modules, they showed the most outstanding performance of the three models tested in this study. The container modules prevent defects by stimulating early rooting in environments that with poor conditions for growth, or in trees that are not suited to the summer environment Therefore, it is expected that they would be an optimal means by which to enable compliance with rules such as the regulation presented by the PPS.

• Proceedings of the KSRS Conference
• /
• v.2
• /
• pp.868-871
• /
• 2006

Mortality of domestic people from cardiovascular disease ranked second, which followed that of from cancer last year. Therefore, it is very important and urgent to enhance the reliability of medical examination and treatment for cardiovascular disease. Heart Rate Variability (HRV) is the most commonly used noninvasive methods to evaluate autonomic regulation of heart rate and conditions of a human heart. In this paper, our aim is to extract a quantitative measure for HRV to enhance the reliability of medical examination for cardiovascular disease, and then develop a prediction method for extracting multi-parametric features by analyzing HRV from ECG. In this study, we propose a hybrid Bayesian classifier called FP-based Bayesian. The proposed classifier use frequent patterns for building Bayesian model. Since the volume of patterns produced can be large, we offer a rule cohesion measure that allows a strong push of pruning patterns in the pattern-generating process. We conduct an experiment for the FP-based Bayesian classifier, which utilizes multiple rules and pruning, and biased confidence (or cohesion measure) and dataset consisting of 670 participants distributed into two groups, namely normal and patients with coronary artery disease.

• Proceedings of the KOR-KST Conference
• /
• 2000.02a
• /
• pp.25-47
• /
• 2000

In route guidance systems fastest-path routing has typically been adopted because of its simplicity. However, empirical studies on route choice behavior have shown that drivers use numerous criteria in choosing a route. The objective of this study is to develop computationally efficient algorithms for identifying a manageable subset of the nondominated (i.e. Pareto optimal) paths for real-time vehicle routing which reflect the drivers' preferences and route choice behaviors. We propose two pruning algorithms that reduce the search area based on a context-dependent linear utility function and thus reduce the computation time. The basic notion of the proposed approach is that ⅰ) enumerating all nondominated paths is computationally too expensive, ⅱ) obtaining a stable mathematical representation of the drivers' utility function is theoretically difficult and impractical, and ⅲ) obtaining optimal path given a nonlinear utility function is a NP-hard problem. Consequently, a heuristic two-stage strategy which identifies multiple routes and then select the near-optimal path may be effective and practical. As the first stage, we utilize the relaxation based pruning technique based on an entropy model to recognize and discard most of the nondominated paths that do not reflect the drivers' preference and/or the context-dependency of the preference. In addition, to make sure that paths identified are dissimilar in terms of links used, the number of shared links between routes is limited. We test the proposed algorithms in a large real-life traffic network and show that the algorithms reduce CPU time significantly compared with conventional multi-criteria shortest path algorithms while the attributes of the routes identified reflect drivers' preferences and generic route choice behaviors well.

• Journal of the Korean Geographical Society
• /
• v.47 no.6
• /
• pp.825-835
• /
• 2012

The aim of this research is to develope technique and mapping for detecting distribution of vegetation dieback areas in the subalpine zone of Mt. Baekdu. A detection technique developed the rule-based model using MODIS images. Dieback areas could be classified as 4 categories of initial dieback, middle dieback, and end dieback by pruning stages of leaves. Dieback area was $28km^2$ from year 2001 to year 2006, intial dieback was $16km^2$, middle dieback was $10km^2$, and end dieback was $2km^2$ by the each stage. Dieback area was $35km^2$ from year 2006 to year 2011. Total area was $35km^2$ from year 2001 to year 2011, areas of middle dieback and end dieback were increased. The research method for this study may help to support in application with preliminary detection of dieback areas in the mountains by the global warming.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.850-858
• /
• 2022

To deploy Gate Recurrent Units (GRU) on resource-constrained embedded devices, this paper presents a reconfigurable FPGA-based GRU accelerator that enables structured compression. Firstly, a dense GRU model is significantly reduced in size by hybrid quantization and structured top-k pruning. Secondly, the energy consumption on external memory access is greatly reduced by the proposed reuse computing pattern. Finally, the accelerator can handle a structured sparse model that benefits from the algorithm-hardware co-design workflows. Moreover, inference tasks can be flexibly performed using all functional dimensions, sequence length, and number of layers. Implemented on the Intel DE1-SoC FPGA, the proposed accelerator achieves 45.01 GOPs in a structured sparse GRU network without batching. Compared to the implementation of CPU and GPU, low-cost FPGA accelerator achieves 57 and 30x improvements in latency, 300 and 23.44x improvements in energy efficiency, respectively. Thus, the proposed accelerator is utilized as an early study of real-time embedded applications, demonstrating the potential for further development in the future.

• Journal of Positioning, Navigation, and Timing
• /
• v.8 no.4
• /
• pp.193-200
• /
• 2019

The auto-encoder network which is a good candidate to handle the modeling of the signal strength attenuation is designed for denoising and compensating the distortion of the received data. It provides a non-linear mapping function by iteratively learning the encoder and the decoder. The encoder is the non-linear mapping function, and the decoder demands accurate data reconstruction from the representation generated by the encoder. In addition, the adaptive network width which supports the automatic generation of new hidden nodes and pruning of inconsequential nodes is also implemented in the proposed algorithm for increasing the efficiency of the algorithm. Simulation results show that the proposed method can improve the neural network training surface to achieve the highest possible accuracy of the signal modeling compared with the conventional modeling method.

• Proceedings of the KIEE Conference
• /
• 2002.07d
• /
• pp.2096-2098
• /
• 2002

In this paper, the optimal evasive maneuver strategies for typical subsonic ASM(anti-ship missile) to reach its target ship with high survivability against CIWS(close in weapon system) are studied. The optimal evasive maneuver input is defined by the homing command optimizing the cost function which takes aiming errors of CIWS into account. The optimization problem for the effective evasive maneuver is formulated based on a simple missile dynamics model and a CIWS model. By means of solving the problem, a multiple hypotheses testing method is proposed. Since this method requires generation of too many hypotheses, the hypothesis-pruning technique is adopted. The solution shows that the optimal evasive maneuver is a bang-bane shaped command whose frequency is varied by the aimpoint determination strategy in CIWS.

• Geomechanics and Engineering
• /
• v.35 no.1
• /
• pp.67-80
• /
• 2023

Due to the nature of the conjunctive Cone Penetration Test(CPT), which does not verify the actual sample directly, geotechnical engineers commonly classify the underground geomaterials using CPT results with the classification diagrams proposed by various researchers. However, such classification diagrams may fail to reflect local geotechnical characteristics, potentially resulting in misclassification that does not align with the actual stratification in regions with strong local features. To address this, this paper presents an objective method for more accurate local CPT soil classification criteria, which utilizes C4.5 decision tree models trained with the CPT results from the clay-dominant southern coast of Korea and the sand-dominant region in South Carolina, USA. The results and analyses demonstrate that the C4.5 algorithm, in conjunction with oversampling, outlier removal, and pruning methods, can enhance and optimize the decision tree-based CPT soil classification model.