• 제어로봇시스템학회:학술대회논문집
• /
• 1993.10b
• /
• pp.109-114
• /
• 1993

It is well known that neural networks can be used to diagnose multiple faults to some limited extent. In this work we present a Multiple Fault Diagnosis Method (MFDM) via neural network which can effectively diagnose multiple faults. To diagnose multiple fault, the proposed method finds the maximum value in the output nodes of the neural network and decreases the node value by changing the hidden node values. This method can find the other faults by computing again with the changed hidden node values. The effectiveness of this method is explored through a neural-network-based fault diagnosis case study of a fluidized catalytic cracking unit (FCCU).

• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.10a
• /
• pp.363-366
• /
• 2001

It is very difficult to maintain the performance of estimation models for the new breed of projects since the computing environment changes so rapidly in terms of programming languages, development tools, and methodologies. So, we propose to use the relevant cases for a neural network model, whose cost is the decreased number of cases. To balance the relevance and data availability, the qualitative input factors are used as criteria of data classification. With the data sets that have the same value for certain qualitative input factors, we can eliminate the factors from the model making reduced neural network models. So we need to seek the optimally reduced neural network model among them. To find the optimally case-selective neural network, we propose the search techniques and sensitivity analysis between data points and search space.

• Journal of information and communication convergence engineering
• /
• v.19 no.3
• /
• pp.148-154
• /
• 2021

With the advent of context-aware computing, many attempts were made to understand emotions. Among these various attempts, Speech Emotion Recognition (SER) is a method of recognizing the speaker's emotions through speech information. The SER is successful in selecting distinctive 'features' and 'classifying' them in an appropriate way. In this paper, the performances of SER using neural network models (e.g., fully connected network (FCN), convolutional neural network (CNN)) with Mel-Frequency Cepstral Coefficients (MFCC) are examined in terms of the accuracy and distribution of emotion recognition. For Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS) dataset, by tuning model parameters, a two-dimensional Convolutional Neural Network (2D-CNN) model with MFCC showed the best performance with an average accuracy of 88.54% for 5 emotions, anger, happiness, calm, fear, and sadness, of men and women. In addition, by examining the distribution of emotion recognition accuracies for neural network models, the 2D-CNN with MFCC can expect an overall accuracy of 75% or more.

• Communications for Statistical Applications and Methods
• /
• v.31 no.1
• /
• pp.105-127
• /
• 2024

We propose recurrent neural network (RNN) methods for forecasting realized volatility (RV). The data are RVs of ten major stock price indices, four from the US, and six from the EU. Forecasts are made for relative ratio of adjacent RVs instead of the RV itself in order to avoid the out-of-scale issue. Forecasts of RV ratios distribution are first constructed from which those of RVs are computed which are shown to be better than forecasts constructed directly from RV. The apparent asymmetry of RV ratio is addressed by the Piecewise Min-max (PM) normalization. The serial dependence of the ratio data renders us to consider two architectures, long short-term memory (LSTM) and gated recurrent unit (GRU). The hyperparameters of LSTM and GRU are tuned by the nested cross validation. The RNN forecast with the PM normalization and ratio transformation is shown to outperform other forecasts by other RNN models and by benchmarking models of the AR model, the support vector machine (SVM), the deep neural network (DNN), and the convolutional neural network (CNN).

• Proceedings of the IEEK Conference
• /
• 2001.06c
• /
• pp.161-164
• /
• 2001

This paper partially presents a Hierachical neural network architecture for providing the intelligent control of complex Automatic Transmission(A/T) system which is usually nonlinear and hard to model mathematically. It consists of the module to apply or release an engine brake at the slope and that to judge the intention of the driver. The HNN architecture simplifies the structure of the overall system and is efficient for the learning time. This paper describes how the sub-neural networks of each module have been constructed and will compare the result of the intelligent A/T control to that of the conventional shift pattern.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.1550-1552
• /
• 2005

In this paper, we have designed a Fingerprint Recognition System based on the Embedded LINUX. The fingerprint is captured using the AS-S2 semiconductor sensor. To extract a feature vector we transform the image of the fingerprint into a column vector. The image is row-wise filtered with the low-pass filter of the Haar wavelet. The feature vectors of the different fingerprints are compared by computing with the probabilistic neural network the distance between the target feature vector and the stored feature vectors in advance. The system implemented consists of a server PC based on the LINUX and a client based on the Embedded LINUX. The client is a Tynux box-x board using a PXA-255 CPU. The algorithm is simple and fast in computing and comparing the fingerprints.

• Proceedings of the KIEE Conference
• /
• 2005.10b
• /
• pp.204-206
• /
• 2005

In this paper, we have designed a Fingerprint Recognition System based on the Embedded LINUX. The fingerprint is captured using the AS-S2 semiconductor sensor. To extract a feature vector we transform the image of t10he fingerprint into a column vector. The image is row-wise filtered with the low-pass filter of the Haar wavelet. The feature vectors of the different fingerprints are compared by computing with the probabilistic neural network the distance between the target feature vector and the stored feature vectors in advance. The system implemented consists of a server PC based on the LINUX and a client based on the Embedded LINUX. The client is a Tynux box-x board using a PXA-255 CPU. The algorithm is simple and fast in computing and comparing the fingerprints.

• Journal of Electrical Engineering and Technology
• /
• v.5 no.3
• /
• pp.363-370
• /
• 2010

Load forecasting has always been essential to the operation and planning of power systems in deregulated electricity markets. Various methods have been proposed for load forecasting, and the neural network is one of the most widely accepted and used techniques. However, to obtain more accurate results, more information is needed as input variables, resulting in huge computational costs in the learning process. In this paper, to reduce training time in multi-layer perceptron-based short-term load forecasting, a graphics processing unit (GPU)-based computing method is introduced. The proposed approach is tested using the Korea electricity market historical demand data set. Results show that GPU-based computing greatly reduces computational costs.

• Electronics and Telecommunications Trends
• /
• v.33 no.6
• /
• pp.24-33
• /
• 2018

In the age of IoT, in which everything is connected to a network, there have been increases in the amount of data traffic, latency, and the risk of personal privacy breaches that conventional cloud computing technology cannot cope with. The idea of edge computing has emerged as a solution to these issues, and furthermore, the concept of ultra-low power edge intelligent semiconductors in which the IoT device itself performs intelligent decisions and processes data has been established. The key elements of this function are an intelligent semiconductor based on artificial intelligence, connectivity for the efficient connection of neurons and synapses, and a large-scale spiking neural network simulation framework for the performance prediction of a neural network. This paper covers the current trends in ultra-low power edge intelligent semiconductors including issues regarding their technology and application.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1056-1059
• /
• 2003

Self-Oranizing Map(SOM) is an unsupervised neural network providing cluster analysis of high dimensional input data. The output from the SOM is represented in map that help us to explore data. The weak point of conventional SOM is when the map is large, it take a long time to train the data. The computing time is known to be O(MN) for trainning to find the winning node (M,N are the number of nodes in width and height of the map). This paper presents a new method to reduce the computing time by creating new map. Each node in a new map is the centroid of nodes' group that are in the original map. After create a new map, we find the winning node of this map, then find the winning node in original map only in nodes that are represented by the winning node from the new map. This new method is called "High Speed Self-Oranizing Map"(HS-SOM). Our experiment use HS-SOM to cluster documents and compare with SOM. The results from the experiment shows that HS-SOM can reduce computing time by 30%-50% over conventional SOM.