• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.11
• /
• pp.1489-1495
• /
• 1995

In this paper, we propose recurrent neural prediction models (RNPM), recurrent neural networks trained as a nonlinear predictor of speech, as a new connectionist model for speech recognition. RNPM modulates its mapping effectively by internal representation, and it requires no time alignment algorithm. Therefore, computational load at the recognition stage is reduced substantially compared with the well known predictive neural networks (PNN), and the size of the required memory is much smaller. And, RNPM does not suffer from the problem of deciding the time varying target function. In the speaker dependent and independent speech recognition experiments under the various conditions, the proposed model was comparable in recognition performance to the PNN, while retaining the above merits that PNN doesn't have.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.56 no.1
• /
• pp.28-36
• /
• 2007

Sensor networks are usually composed of tens or thousands of tiny devices with limited resources. Because of their limited resources, many researchers have studied on the energy management in the WSNs(Wireless Sensor Networks), especially taking into account communications efficiency. For effective data transmission and sensor fault detection in sensor network environment, a new remote monitoring system based on PCA(Principle Component Analysis) and AANN(Auto Associative Neural Network) is proposed. PCA and AANN have emerged as a useful tool for data compression and identification of abnormal data. Proposed system can be effectively applied to sensor network working in LEA2C(Low Energy Adaptive Connectionist Clustering) routing algorithms. To verify its applicability, some simulation studies on the data obtained from real WSNs are executed.

• Proceedings of the KAIS Fall Conference
• /
• 2008.05a
• /
• pp.152-156
• /
• 2008

We consider the problem of constructing observers for nonlinear systems with unknown inputs. Connectionist networks, also called neural networks, have been broadly applied to solve many different problems since McCulloch and Pitts had shown mathematically their information processing ability in 1943. In this thesis, we present a genetic neuro-control scheme for nonlinear systems. Our method is different from those using supervised learning algorithms, such as the backpropagation (BP) algorithm, that needs training information in each step. The contributions of this thesis are the new approach to constructing neural network architecture and its training.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.29.1-29
• /
• 2001

Neuro-fuzzy systems are multi-layered connectionist networks that realize the elements and functions of traditional fuzzy logic control/decision systems. A trained neuro-fuzzy system is isomorphic to a fuzzy logic system, and fuzzy IF-THEN rule knowledge can be explicitly extracted from the network. This talk presents a brief introduction to self-adaptive neuro-fuzzy systems and addresses some recent research results and applications. Most of the existing neuro-fuzzy systems exhibit several major drawbacks that lead to performance degradation. These drawbacks are the curse of dimensionality (i.e., fuzzy rule explosion), inability to re-structure their internal nodes in a changing environment, and their lack of ability to extract knowledge from a given set of training data. This talk focuses on our investigation of network architectures, self-adaptation algorithms, and efficient learning algorithms that will enable existing neuro-fuzzy systems to self-adapt themselves in an unstructured and uncertain environment.

• Korean Journal of Cognitive Science
• /
• v.23 no.2
• /
• pp.133-164
• /
• 2012

Paul Churchland(1989) suggests the theory of representation from the results of cognitive biology and connectionist AI studies. According to the theory, our representations of the diverse phenomena in the world can be represented as the positions of phase state spaces with the actions of the neurons or of the assembly of neurons. He insists connectionist AI neural networks can have the semantical category systems to recognize the world. But Fodor and Lepore(1996) don't look the perspective bright. From their points of view, the Churchland's theory of representation stands on the base of Quine's holism, and the network semantics cannot explain how the criteria of semantical content similarity could be possible, and so cannot the theory. This thesis aims to excavate which one is the better between the perspective of the theory and the one of Fodor and Lepore's. From my understandings of state space theory of representation, artificial nets can coordinates the criteria of contents similarity by the learning algorithm. On the basis of these, I can see that Fodor and Lepore's points cannot penetrate the Churchlands' theory. From the view point of the theory, we can see how the future's artificial systems can have the conceptual systems recognizing the world. Therefore we can have the perspectives what cognitive scientists have to focus on.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.3
• /
• pp.115-122
• /
• 2019

For Korean phoneme recognition, Hidden Markov-Gaussian Mixture model(HMM-GMM) or hybrid models which combine artificial neural network with HMM have been mainly used. However, current approach has limitations in that such models require force-aligned corpus training data that is manually annotated by experts. Recently, researchers used neural network based phoneme recognition model which combines recurrent neural network(RNN)-based structure with connectionist temporal classification(CTC) algorithm to overcome the problem of obtaining manually annotated training data. Yet, in terms of implementation, these RNN-based models have another difficulty in that the amount of data gets larger as the structure gets more sophisticated. This problem of large data size is particularly problematic in the Korean language, which lacks refined corpora. In this study, we introduce CTC algorithm that does not require force-alignment to create a Korean phoneme recognition model. Specifically, the phoneme recognition model is based on convolutional neural network(CNN) which requires relatively small amount of data and can be trained faster when compared to RNN based models. We present the results from two different experiments and a resulting best performing phoneme recognition model which distinguishes 49 Korean phonemes. The best performing phoneme recognition model combines CNN with 3hop Bidirectional LSTM with the final Phoneme Error Rate(PER) at 3.26. The PER is a considerable improvement compared to existing Korean phoneme recognition models that report PER ranging from 10 to 12.

• Journal of the Korea Society of Computer and Information
• /
• v.6 no.2
• /
• pp.48-57
• /
• 2001

In a fuzzy connectionist expert system(FCES), the knowledge base can be constructed of neural logic networks to represent fuzzy rules and their relationship, We call it fuzzy rule inference network. To find out the belief value of a conclusion, the traditional inference strategy in a FCES will back-propagate from a rule term of the conclusion and follow through the entire network sequentially This sequential search strategy is very inefficient. In this paper, to improve the above search strategy, we proposed fuzzy rule inference rule used in a FCES was modified. The proposed adaptive search strategy in fuzzy rule inference network searches the network according to the search priorities.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09b
• /
• pp.95-98
• /
• 2003

In this paper, an adaptive fuzzy neural control of unknown nonlinear systems based on the rapid learning algorithm is proposed for optimal parameterization. We combine the advantages of fuzzy control and neural network techniques to develop an adaptive fuzzy control system for updating nonlinear parameters of controller. The Fuzzy Neural Network(FNN), which is constructed by an equivalent four-layer connectionist network, is able to learn to control a process by updating the membership functions. The free parameters of the AFN controller are adjusted on-line according to the control law and adaptive law for the purpose of controlling the plant track a given trajectory and it's initial values are off-line preprocessing, In order to improve the convergence of the learning process, we propose a rapid learning algorithm which combines the error back-propagation algorithm with Aitken's $\delta$$\^$2/ algorithm. The heart of this approach ls to reduce the computational burden during the FNN learning process and to improve convergence speed. The simulation results for nonlinear plant demonstrate the control effectiveness of the proposed system for optimal parameterization.

• The KIPS Transactions:PartB
• /
• v.13B no.4 s.107
• /
• pp.421-428
• /
• 2006

Intelligent Tutoring System(ITS), which offers individualized learning environment that consider many learners' variable, is realized by the effective alternative to take the place of domain expert. Accordingly, research on Learning Companion System(LC) is currently noticing. However, to develop LCS which applies effective interaction, it is necessary to combine several LCs, and personalized knowledge base have to be made first. Therefore, in this paper, we propose the 'Knowledge Base Object Medel', which is based on connectionist' in cognition structure, represents learner's knowledge to self-learnig object, and grows adaptive object by proprietor, verify the validity. This model lays the groundwork for design of personalized knowledge base, offers clue to development of adaptive ITS using knowledge base object.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2001.10a
• /
• pp.535-538
• /
• 2001

Since 1990s, many literatures have shown that connectionist models, such as back propagation, recurrent network, and RBF (Radial Basis Function) outperform the traditional models, MA (Moving Average), AR (Auto Regressive), and ARIMA (Auto Regressive Integrated Moving Average) in time series prediction. Neural based approaches to time series prediction require the enough length of historical measurements to generate the enough number of training patterns. The more training patterns, the better the generalization of MLP is. The researches about the schemes of generating artificial training patterns and adding to the original ones have been progressed and gave me the motivation of developing VTG schemes in 1996. Virtual term is an estimated measurement, X(t+0.5) between X(t) and X(t+1), while the given measurements in the series are called actual terms. VTG (Virtual Tern Generation) is the process of estimating of X(t+0.5), and VTG schemes are the techniques for the estimation of virtual terms. In this paper, the alternative VTG schemes to the VTG schemes proposed in 1996 will be proposed and applied to multivariate time series prediction. The VTG schemes proposed in 1996 are called deterministic VTG schemes, while the alternative ones are called stochastic VTG schemes in this paper.