• International Journal of Control, Automation, and Systems
• /
• v.1 no.3
• /
• pp.321-331
• /
• 2003

Experimental software datasets describing software projects in terms of their complexity and development time have been the subject of intensive modeling. A number of various modeling methodologies and modeling designs have been proposed including such approaches as neural networks, fuzzy, and fuzzy neural network models. In this study, we introduce the concept of the Rule-based fuzzy polynomial neural networks (RFPNN) as a hybrid modeling architecture and discuss its comprehensive design methodology. The development of the RFPNN dwells on the technologies of Computational Intelligence (CI), namely fuzzy sets, neural networks, and genetic algorithms. The architecture of the RFPNN results from a synergistic usage of RFNN and PNN. RFNN contribute to the formation of the premise part of the rule-based structure of the RFPNN. The consequence part of the RFPNN is designed using PNN. We discuss two kinds of RFPNN architectures and propose a comprehensive learning algorithm. In particular, it is shown that this network exhibits a dynamic structure. The experimental results include well-known software data such as the NASA dataset concerning software cost estimation and the one describing software modules of the Medical Imaging System (MIS).

• Management Science and Financial Engineering
• /
• v.3 no.2
• /
• pp.29-43
• /
• 1997

Many heuristics have been developed in order to overcome the computational complexity of job shop problems. In this research, we develop a new heuristic by selecting four simple dispatching rules, i.e., SPT, LPT, SR and LR, dynamically as scheduling proceeds. The selection is accomplished by using artificial neural networks. As a result of testing on 50 problems, the makespan obtained by our heuristic is, on the average, 13.0% shorter than the longest makespan, and 0.4% shorter than the shortest makespan obtained by existing dispatching rules.

• Proceedings of the IEEK Conference
• /
• 2002.06c
• /
• pp.87-90
• /
• 2002

This paper has been studied a new approach using neural network to detect false laser marks. In the proposed approach, input images are segmented into R, G and B colors and implements mask areas respectively. And then average and variation values of the each mask area are extracted for the learning process to minimize input nodes. Using this technique, the new input data is obtained and implemented to the back-propagation algorithm using multi layer perception. This paper reduces the computational complexity necessary and shows better effectiveness to inspect false laser marks.

• Speech Sciences
• /
• v.11 no.4
• /
• pp.53-66
• /
• 2004

The nonlinear speech enhancement method with interactive parallel-extended Kalman filter is applied to speech contaminated by additive white noise. To represent the nonlinear and nonstationary nature of speech. we assume that speech is the output of a nonlinear prediction HMM (NPHMM) combining both neural network and HMM. The NPHMM is a nonlinear autoregressive process whose time-varying parameters are controlled by a hidden Markov chain. The simulation results shows that the proposed method offers better performance gains relative to the previous results [6] with slightly increased complexity.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.7 no.2
• /
• pp.132-137
• /
• 2007

Experimental software data capturing the essence of software projects (expressed e.g., in terms of their complexity and development time) have been a subject of intensive modeling. In this study, we introduce a new category of Hybrid Fuzzy Neural Networks (gHFNN) and discuss their comprehensive design methodology. The gHFNN architecture results from highly synergistic linkages between Fuzzy Neural Networks (FNN) and Polynomial Neural Networks (PNN). We develop a rule-based model consisting of a number of "if-then" statements whose antecedents are formed in the input space and linked with the consequents (conclusion pats) formed in the output space. In this framework, FNNs contribute to the formation of the premise part of the overall network structure of the gHFNN. The consequences of the rules are designed with the aid of genetically endowed PNNs. The experiments reported in this study deal with well-known software data such as the NASA dataset. In comparison with the previously discussed approaches, the proposed self-organizing networks are more accurate and yield significant generalization abilities.

• Journal of the Korea Society of Computer and Information
• /
• v.16 no.6
• /
• pp.61-70
• /
• 2011

In this paper, it is realized an image filter for a noise elimination using a recurrent neural networks with fuzzy. The proposed fuzzy neural networks structure is to converge weights and the number of iteration for a certain value by using basically recurrent neural networks structure and is simplified computation and complexity of mathematics by applying the hybrid fuzzy membership function operator. In this paper, the proposed method, the recurrent neural networks applying fuzzy which is collected a certain value, has been proved improving average 0.38dB than the conventional method, the generalied recurrent neural networks, by using PSNR. Also, a result image of the proposed method was similar to the original image than a result image of the conventional method by comparing to visual images.

• ETRI Journal
• /
• v.42 no.4
• /
• pp.505-517
• /
• 2020

The increasing size and complexity of deep neural networks (DNNs) necessitate the development of efficient high-performance accelerators. An efficient memory structure and operating scheme provide an intuitive solution for high-performance accelerators along with dataflow control. Furthermore, the processing of various neural networks (NNs) requires a flexible memory architecture, programmable control scheme, and automated optimizations. We first propose an efficient architecture with flexibility while operating at a high frequency despite the large memory and PE-array sizes. We then improve the efficiency and usability of our architecture by automating the optimization algorithm. The experimental results show that the architecture increases the data reuse; a diagonal write path improves the performance by 1.44× on average across a wide range of NNs. The automated optimizations significantly enhance the performance from 3.8× to 14.79× and further provide usability. Therefore, automating the optimization as well as designing an efficient architecture is critical to realizing high-performance DNN accelerators.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.2498-2501
• /
• 2003

In this paper, we present a wavelet neural network (WNN) approach to the solution of the path tracking problem for mobile robots that possess complexity, nonlinearity and noise. First, we discuss a WNN based control system where the control signals are directly obtained by minimizing the difference between the reference track and the pose of a mobile robot. This compact network structure is helpful to determine the number of hidden nodes and the initial value of weights. Then, the data with various noises provided by odometric and external sensors are here fused together by means of an Extended Kalman Filter (EKF) approach for the pose estimation problem of mobile robots. This control process is a dynamic on-line process that uses the wavelet neural network trained via the gradient-descent method with estimates from EKF. Finally, we verify the effectiveness and feasibility of the proposed control system through simulations.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2005.09a
• /
• pp.167-172
• /
• 2005

This paper introduces a method which improves the performance of the identification of splice sites in the genomic DNA sequence of eukaryotes. This method combines a low order Markov model in series with a neural network for the predictions of splice sites. The lower order Markov model incorporates the biological knowledge surrounding the splice sites as probabilistic parameters. The Neural network takes the Markov encoded parameters as the inputs and produces the prediction. Two types of neural networks are used for the comparison. This method reduces the computational complexity and shows encouraging accuracy in the predictions of splice sites when applied to several standard splice site dataset.

• Journal of Ocean Engineering and Technology
• /
• v.20 no.5 s.72
• /
• pp.70-76
• /
• 2006

A Probabilistic neural network (PNN) technique for predicting the stability number for the armor blocks of breakwaters is presented. A PNN is prepared using the experimental data of van der Meer and is then compared with the empirical formula and previous artificial neural network (ANN) model. This comparison shows that a PNN can effectively predict the stability numbers in spite of data complexity, incompleteness, and incoherence, and can be an effective tool for the designers of rubble mound breakwaters to support their decision process and to improve design efficiency.