• Proceedings of the KIEE Conference
• /
• 1997.07b
• /
• pp.739-742
• /
• 1997

This paper proposes the structure and parameter optimization technique of fuzzy neural networks using genetic algorithm. Fuzzy neural network has advantages of both the fuzzy inference system and neural network. The determination of the optimal parameters and structure of the fuzzy neural networks, however, requires special efforts. To solve these problems, we propose a new learning method for optimization of fuzzy neural networks using genetic algorithm. It can optimize the structure and parameters of the entire fuzzy neural network globally. Numerical example is provided to show the advantages of the proposed method.

• Wind and Structures
• /
• v.22 no.5
• /
• pp.543-553
• /
• 2016

The growing complexity of modern technological systems requires more flexible and powerful reliability analysis tools. Existing tools encounter a number of limitations including lack of modeling power to address components interactions for complex systems and lack of flexibility in handling component failure distribution. We propose a reliability modeling framework based on the Bayesian network (BN). It can combine historical data with expert judgment to treat data scarcity. The proposed methodology is applied to wind turbines reliability analysis. The observed result shows that a BN based reliability modeling is a powerful potential solution to modeling and analyzing various kinds of system components behaviors and interactions. Moreover, BN provides performing several inference approaches such as smoothing, filtering, what-if analysis, and sensitivity analysis for considering system.

• Journal of Korea Multimedia Society
• /
• v.25 no.5
• /
• pp.676-684
• /
• 2022

As realistic media are widespread in various image processing areas, image or video compression is one of the key technologies to enable real-time applications with limited network bandwidth. Generally, image or video compression cause the unnecessary compression artifacts, such as blocking artifacts and ringing effects. In this study, we propose a Deep Residual Channel-attention Network, so called DRCAN, which consists of an input layer, a feature extractor and an output layer. Experimental results showed that the proposed DRCAN can reduced the total memory size and the inference time by as low as 47% and 59%, respectively. In addition, DRCAN can achieve a better peak signal-to-noise ratio and structural similarity index measure for compressed images compared to the previous methods.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.410-412
• /
• 1998

Subsymbolic Knowledge processing is said to be changed states of networks constructed from small elements. subsymbolic systems also make it possible to use connectionist models for knowledge processing. Connectionist realization such modulus are modulus linked together for solving a given problem. We study using neural networks as distinct actions. The output vectors produced by the neural networks are consider as a new facts. These new facts are then processed to activate another networks or used in the current production rule, The production rule is applying knowledge stored in the knowledge base to make inference. After neural networks knowledge base is constructed and trained. We present a running sample of incorporating neural network knowledge base. We implement using rochester connectionist simulator. We suggest that incorporating neural network knowledge base. Therefore incorporated neural network knowledge base ensures a cleaner solution which results in better perfor s.

• International Journal of Computer Science & Network Security
• /
• v.22 no.12
• /
• pp.1-12
• /
• 2022

Besides unexpected growth perceived by IoT's, the variety and volume of threats have increased tremendously, making it a necessity to introduce intrusion detections systems for prevention and detection of such threats. But Intrusion Detection and Prevention System (IDPS) inside the IoT network yet introduces some unique challenges due to their unique characteristics, such as privacy inference, performance, and detection rate and their frequency in the dynamic networks. Our research is focused on the privacy inferences of existing intrusion prevention and detection system approaches. We also tackle the problem of providing unified a solution to implement the open-source IDPS in the IoT architecture for assessing the performance of IDS by calculating; usage consumption and detection rate. The proposed scheme is considered to help implement the human health monitoring system in IoT networks

• Electronics and Telecommunications Trends
• /
• v.38 no.5
• /
• pp.12-22
• /
• 2023

The development of neural networks is evolving towards the adoption of transformer structures with attention modules. Hence, active research focused on extending the concept of lightweight neural network algorithms and hardware acceleration is being conducted for the transition from conventional convolutional neural networks to transformer-based networks. We present a survey of state-of-the-art research on lightweight neural network algorithms and hardware architectures to reduce memory usage and accelerate both inference and training. To describe the corresponding trends, we review recent studies on token pruning, quantization, and architecture tuning for the vision transformer. In addition, we present a hardware architecture that incorporates lightweight algorithms into artificial intelligence processors to accelerate processing.

• Journal of the Korean Society for Precision Engineering
• /
• v.14 no.9
• /
• pp.116-123
• /
• 1997

We have already presented the constraint satisfaction algorithm which could solve the closed loop porblem in constraint network by using local constraint propagation, variable elimination and constraint modularization. With this algorithm, we have implemented a knowledge-based system (intelligent CAD) for supporting machine design interactively. In this paper, we present newer constraint satisfaction algorithm which can solve inequalities or under-constrained problems in constraint network, interactively and effi- ciently. This algorithm is a hybrid type of using both declarative description (constraint representation) and optimization algorithm (Simulated Annealing), simultaneously. The under-constrained problems are represented by constraint networks and satisfied completely with this algorithm. The usefulness of our algorithm will be illustrated by the application to a gear design.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2005.09a
• /
• pp.202-205
• /
• 2005

Cell phenotypes are determined by the concerted activity of thousands of genes and their products. This activity is coordinated by a complex network that regulates the expression of genes. Understanding this organization is crucial to elucidate cellular activities, and many researches have tried to construct gene regulatory networks from mRNA expression data which are nowadays the most available and have a lot of information for cellular processes. Several computational tools, such as Boolean network, Qualitative network, Bayesian network, and so on, have been applied to infer these networks. Among them, Bayesian networks that we chose as the inference tool have been often used in this field recently due to their well-established theoretical foundation and statistical robustness. However, the relative insufficiency of experiments with respect to the number of genes leads to many false positive inferences. To alleviate this problem, we had developed the algorithm of MONET(MOdularized NETwork learning), which is a new method for inferring modularized gene networks by utilizing two complementary sources of information: biological annotations and gene expression. Afterward, we have packaged and improved MONET by combining dispersed functional blocks, extending species which can be inputted in this system, reducing the time complexities by improving algorithms, and simplifying input/output formats and parameters so that it can be utilized in actual fields. In this paper, we present the architecture of MONET system that we have improved.

• Journal of Pharmacopuncture
• /
• v.18 no.3
• /
• pp.11-18
• /
• 2015

Objectives: Systems biology is a novel subject in the field of life science that aims at a systems' level understanding of biological systems. Because of the significant progress in high-throughput technologies and molecular biology, systems biology occupies an important place in research during the post-genome era. Methods: The characteristics of systems biology and its applicability to traditional medicine research have been discussed from three points of view: data and databases, network analysis and inference, and modeling and systems prediction. Results: The existing databases are mostly associated with medicinal herbs and their activities, but new databases reflecting clinical situations and platforms to extract, visualize and analyze data easily need to be constructed. Network pharmacology is a key element of systems biology, so addressing the multi-component, multi-target aspect of pharmacology is important. Studies of network pharmacology highlight the drug target network and network target. Mathematical modeling and simulation are just in their infancy, but mathematical modeling of dynamic biological processes is a central aspect of systems biology. Computational simulations allow structured systems and their functional properties to be understood and the effects of herbal medicines in clinical situations to be predicted. Conclusion: Systems biology based on a holistic approach is a pivotal research methodology for understanding the mechanisms of traditional medicine. If systems biology is to be incorporated into traditional medicine, computational technologies and holistic insights need to be integrated.

• Proceedings of the Korean Geotechical Society Conference
• /
• 2009.03a
• /
• pp.456-463
• /
• 2009

The preconsolidation pressure has been commonly determined by oedometer test. However, it can also be determined by in-situ test, such as piezocone test with theoretical and(or) empirical correlations. Recently, Neural Network(NN) theory was applied and some models were proposed to estimate the preconsolidation pressure or OCR. However, since the optimization process of synaptic weights of NN model is dependent on the initial synaptic weights, NN models which are trained with different initial weights can't avoid the variability on prediction result for new database even though they have same structure and use same transfer function. In this study, Committee Neural Network(CNN) model is proposed to improve the initial weight dependency of multi-layered neural network model on the prediction of preconsolidation pressure of soft clay from piezocone test result. It was found that even though the NN model has the optimized structure for given training data set, it still has the initial weight dependency, while the proposed CNN model can improve the initial weight dependency of the NN model and provide a consistent and precise inference result than existing NN models.