• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.3
• /
• pp.534-540
• /
• 2008

The adaptive learning circuit is designed on the basis of modeling of MFSFET (Metal-Ferroelectric-Semiconductor FET) and the numerical results is analyzed. The output frequency of the adaptive learning circuit is inversely proportional to the source-drain resistance of MFSFET and the capacitance of the circuit. The saturated drain current with input pulse number is analogous to the ferroelectric polarization reversal. It indicates that the ferroelectric polarization plays an important role in the drain current control of MFSFET. The output frequency modulation of the adaptive learning circuit is investigated by analyzing the source-drain resistance of MFSFET as functions of input pulse numbers in the adaptive learning circuit and the dimensionality factor of the ferroelectric thin film. From the results, adaptive learning characteristics which means a gradual frequency change of output pulse with the progress of input pulse, are confirmed. Consequently it is shown that our circuit can be used effectively in the neuron synapses of neural networks.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.36S no.7
• /
• pp.81-89
• /
• 1999

Image obtained by incomplete communication always include noise, blur and distortion, etc. In this paper, we propose and apply the new spatial filter algorithm, called an optimal adaptive generalized stack filter(AGSF), which optimizes adaptive generalized stack filter(AGSF) using neural network weight learning algorithm of back-propagation learning algorithm for improving noise removal and edge preservation rate. AGSF divides into two parts: generalized stack filter(GSF) and adaptive multistage median filter(AMMF), GSF improves the ability of stack filter algorithm and AMMF proposes the improved algorithm for reserving the sharp edge. Applied to neural network theory, the proposed algorithm improves the performance of the AGSF using two weight learning algorithms, such as the least mean absolute(LAM) and least mean square (LMS) algorithms. Simulation results of the proposed filter algorithm are presented and discussed.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.05a
• /
• pp.761-763
• /
• 2008

최근 온라인교육의 필요성이 높아지고 요구 수준이 커짐에 따라 교육 서비스를 제공하는 시스템의 지능화된 처리능력이 필요하다. 퍼지신경회로망은 각각의 가중치(weight)를 갖는 채널로 연결한 망형태의 계산모델이다. 퍼지신경회로망을 학습시스템에 적용하여 학습자의 문항테스트 결과에서 학습과정을 재설정 할 수 있는 출력 값을 생성한다. 적응형 학습시스템은 퍼지신경회로망을 적용하여 개별화된 강의 코스로 학습을 진행하고 결과의 feedback을 통해 학습자의 최적 커리큘럼을 찾아내는 방법을 구현하였다.

• Journal of KIISE:Software and Applications
• /
• v.32 no.5
• /
• pp.438-448
• /
• 2005

This paper proposes an adaptive tutoring system that provides learning materials dynamically according to the learners' teaming character and ability. Our system, in which a learning phase and a test phase are linked together, supports the personalized instruction-learning by providing the teaming materials by level in the learning phase according to the teaming ability estimated in the test phase. We design and implement a tutoring system consisted of an evaluation component and a learning component. An evaluation component uses a computerized adaptive test(CAT) based on item response theory to evaluate learners' ability while a learning component employs fuzzy level set theory so that teaming contents are provided to learners according to learners' level.

• Journal of KIISE:Computing Practices and Letters
• /
• v.9 no.4
• /
• pp.431-438
• /
• 2003

In this study, we have developed a web-based adaptive testing system using item response theory´s computerized adaptive testing to diagnose underachievers, and to check the evaluation results immediately. Adaptive testing system simple is not the fact that it presents a question to students. It calculates information of a question and presents the question to students. It controls the response of the students under extraction conditions of the next question. It extracts the question which is the most suitable it presents. In this adaptive testing system, you can extract questions according to the level of the students, and adjust the length and the level of the difficulty according to the response of the students.

• Journal of KIISE:Software and Applications
• /
• v.35 no.1
• /
• pp.33-40
• /
• 2008

In this paper, we propose a method to enhance adaptability in a dialogue system using the reinforcement learning that reduces response errors by trials and error-search similar to a human dialogue process. The adaptive dialogue strategy means that the dialogue system improves users' satisfaction and dialogue efficiency by loaming users' dialogue styles. To apply the reinforcement learning to the dialogue system, we use a main-dialogue span and sub-dialogue spans as the mathematic application units, and evaluate system usability by using features; success or failure, completion time, and error rate in sub-dialogue and the satisfaction in main-dialogue. In addition, we classify users' groups into beginners and experts to increase users' convenience in training steps. Then, we apply reinforcement learning policies according to users' groups. In the experiments, we evaluated the performance of the proposed method on the individual reinforcement learning policy and group's reinforcement learning policy.

• Journal of KIISE
• /
• v.43 no.6
• /
• pp.686-691
• /
• 2016

In this paper, we introduced the adaptive spectral co-clustering, a spectral clustering for multiview data, especially data with more than three views. In the adaptive spectral co-clustering, the performance is improved by sharing information from diverse views. For the efficiency in information sharing, a co-training approach is adopted. In the co-training step, a set of parameters are estimated to make all views in data maximally independent, and then, information is shared with respect to estimated parameters. This co-training step increases the efficiency of information sharing comparing with ordinary feature concatenation and co-training methods that assume the independence among views. The adaptive spectral co-clustering was evaluated with synthetic dataset and multi lingual document dataset. The experimental results indicated the efficiency of the adaptive spectral co-clustering with the performances in every iterations and similarity matrix generated with information sharing.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1997.11a
• /
• pp.176-179
• /
• 1997

본 논문은 입/출력 관계가 불명확한 가전제품 제어에 인공신경회로망을 응용하여 지능형 제어기를 구현하는 방법에 관한 것이다. 다층신경회로망을 사용하고 Error Back Propagation 학습방법에 의하여 학습되도록 한다. 제어대상물에서 알 수 있는 정보는 입력값과 이에 대응하는 출력값 뿐이며 입력과 출력에 대한 관계를 수학적으로 모델링하기 어려운 경우이다. 인공신경회로망을 이용한 제어를 위하여 Neural Network Emulator(NNE)와 Neural Network Controller(NNC)가 개발되며 각 신경회로망의 초기하중백터는 제어대상에 오프라인 학습으로 결정하고, 자동조절과정에서 온라인 학습하여 새로운 대상제품 상황에 적응하도록 설계되었다. 제안된 지능형 제어시스템은 PC를 이용하여 실시스템에 적용하여 검토되었다.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.42 no.6
• /
• pp.37-46
• /
• 2005

In the Web environment with a huge amount of information, Web page access patterns for the users visiting certain web site can be diverse and change continually in accordance with the change of its environment. Therefore it is almost impossible to develop and design web sites which fit perfectly for every web user's desire. Adaptive web site was proposed as solution to this problem. In this paper, we will present an effective method that uses a probabilistic model of DTMC(Discrete-Time Markov Chain) for learning user's access patterns and applying these patterns to construct an adaptive web site.

• Journal of KIISE:Computing Practices and Letters
• /
• v.13 no.5
• /
• pp.338-342
• /
• 2007

In this paper we propose a user adaptive post-processing method to improve the accuracy of speaker dependent, isolated word speech recognition, particularly for mobile devices. Our method considers the recognition result of the basic recognizer simply as a high-level speech feature and processes it further for correct recognition result. Our method learns correlation between the output of the basic recognizer and the correct final results and uses it to correct the erroneous output of the basic recognizer. A multi-layer perceptron model is built for each incorrectly recognized word with high frequency. As the result of experiments, we achieved a significant improvement of 41% in recognition accuracy (41% error correction rate).