• 정보과학회 컴퓨팅의 실제 논문지
• /
• 제22권3호
• /
• pp.127-132
• /
• 2016

인공신경망을 압축에 적용하여 더 높은 압축 성능을 보이기 위한 알고리즘들이 몇 가지 연구되어 있다. 그러나 그동안 이러한 알고리즘들은 한정된 계산 능력의 하드웨어를 가지고 있기에 작은 크기의 신경망을 사용할 수밖에 없었으며 적용하는 대상 역시 실제로 사용하기에는 너무 작은 크기의 파일들이었다. 본 논문에서는 GPGPU의 계산능력을 신경망 학습에 이용하여 만든 텍스트 문맥 기반 문자 등장 확률 예측기와 함께 허프만 부호화의 성능을 높일 수 있는 변환 방법을 제시한다. 앞먹임 신경망과 GRU 회귀 신경망에 대해 실험을 수행하였으며, 회귀 신경망 모델은 앞먹임 신경망에 비해 뛰어난 예측 성공률과 압축률을 보였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2000년도 제15차 학술회의논문집
• /
• pp.494-494
• /
• 2000

This paper demonstrates that the largest Lyapunov exponent $\lambda$ of recurrent neural networks can be controlled by a gradient method. The method minimizes a square error $e_{\lambda}=(\lambda-\lambda^{obj})^2$ where $\lambda^{obj}$ is desired exponent. The $\lambda$ can be given as a function of the network parameters P such as connection weights and thresholds of neurons' activation. Then changes of parameters to minimize the error are given by calculating their gradients $\partial\lambda/\partialP$. In a previous paper, we derived a control method of $\lambda$via a direct calculation of $\partial\lambda/\partialP$ with a gradient collection through time. This method however is computationally expensive for large-scale recurrent networks and the control is unstable for recurrent networks with chaotic dynamics. Our new method proposed in this paper is based on a stochastic relation between the complexity $\lambda$ and parameters P of the networks configuration under a restriction. Then the new method allows us to approximate the gradient collection in a fashion without time evolution. This approximation requires only $O(N^2)$ run time while our previous method needs $O(N^{5}T)$ run time for networks with N neurons and T evolution. Simulation results show that the new method can realize a "stable" control for larege-scale networks with chaotic dynamics.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 1998년도 하계종합학술대회논문집
• /
• pp.263-266
• /
• 1998

This paper propose a prefilter type inverse control algorithm using chaotic neural networks. Since the chaotic neural networks show robust characteristics in approximation and adaptive learning for nonlinear dynamic system, the chaotic neural networks are suitable for controlling robotic manipulators. The structure of the proposed prefilter type controller compensate velocity of the PD controller. To estimate the proposed controller, we implemented to the Cartesian space control of three-axis PUMA robot and compared the final result with recurrent neural network(RNN) controller.

• Transactions on Control, Automation and Systems Engineering
• /
• 제4권2호
• /
• pp.124-129
• /
• 2002

This paper demonstrates that the largest Lyapunov exponent λ of recurrent neural networks can be controlled efficiently by a stochastic gradient method. An essential core of the proposed method is a novel stochastic approximate formulation of the Lyapunov exponent λ as a function of the network parameters such as connection weights and thresholds of neural activation functions. By a gradient method, a direct calculation to minimize a square error (λ - λ$\^$obj/)$^2$, where λ$\^$obj/ is a desired exponent value, needs gradients collection through time which are given by a recursive calculation from past to present values. The collection is computationally expensive and causes unstable control of the exponent for networks with chaotic dynamics because of chaotic instability. The stochastic formulation derived in this paper gives us an approximation of the gradients collection in a fashion without the recursive calculation. This approximation can realize not only a faster calculation of the gradient, but also stable control for chaotic dynamics. Due to the non-recursive calculation. without respect to the time evolutions, the running times of this approximation grow only about as N$^2$ compared to as N$\^$5/T that is of the direct calculation method. It is also shown by simulation studies that the approximation is a robust formulation for the network size and that proposed method can control the chaos dynamics in recurrent neural networks efficiently.

• 한국전기전자재료학회:학술대회논문집
• /
• 한국전기전자재료학회 2004년도 하계학술대회 논문집 Vol.5 No.2
• /
• pp.1234-1238
• /
• 2004

A Controlled neural networks are proposed in order to measure nonlinear environments in adaptive and in realtime. The structure of it is similar to recurrent neural networks: a delayed output as the input and a delayed error between tile output of plant and neural networks as a bias input. In addition, we compute the desired value of hidden layer by an optimal method instead of transfering desired values by backpropagation and each weights are updated by RLS(Recursive Least Square). Consequently, this neural networks are not sensitive to initial weights and a learning rate, and have a faster convergence rate than conventional neural networks. This new neural networks is Error Estimated Neural Networks. We can estimate nonlinear models in realtime by the proposed networks and control nonlinear models. To show the performance of this one, we have various experiments. And this controller call prove effectively to be control in the environments of various systems.

• 한국전기전자재료학회:학술대회논문집
• /
• 한국전기전자재료학회 2008년도 하계학술대회 논문집 Vol.9
• /
• pp.525-526
• /
• 2008

For measuring nonlinear Vehicle Modeling based on 7-Depth Sensor, the neural networks are proposed m adaptive and in realtime. The structure of it is similar to recurrent neural networks; a delayed output as the input and a delayed error between the output of plant and neural networks as a bias input. In addition, we compute the desired value of hidden layer by an optimal method instead of transfering desired values by backpropagation and each weights are updated by RLS(Recursive Least Square). Consequently, this neural networks are not sensitive to initial weights and a learning rate, and have a faster convergence rate than conventional neural networks. This new neural networks is Error Estimated Neural Networks. We can estimate nonlinear models in realtime by the proposed networks and control nonlinear models.

• 한국지능시스템학회논문지
• /
• 제8권2호
• /
• pp.50-59
• /
• 1998

This paper presents stabilization and adaptive control of flexible single link robot manipulator system by self-recurrent neural networks that is one of the neural networks and is effective in nonlinear control. The architecture of neural networks is a modified model of self-recurrent structure which has a hidden layer. The self-recurrent neural networks can be used to approximate any continuous function to any desired degree of accuracy and the weights are updated by feedback-error learning algorithm. When a flexible manipulator is rotated by a motor through the fixed end, transverse vibration may occur. The motor toroque should be controlled in such a way that the motor rotates by a specified angle, while simultaneously stabilizing vibration of the flexible manipuators so that it is arresed as soon as possible at the end of rotation. Accurate vibration control of lightweight manipulator during the large changes in configuration common to robotic tasks requires dynamic models that describe both the rigid body motions, as well as the flexural vibrations. Therefore, a dynamic models for a flexible single link robot manipulator is derived, and then a comparative analysis was made with linear controller through an simulation and experiment. The results are proesented to illustrate thd advantages and imporved performance of the proposed adaptive control ove the conventional linear controller.

• 전자공학회논문지B
• /
• 제32B권11호
• /
• 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.

• 실천공학교육논문지
• /
• 제14권3호
• /
• pp.491-498
• /
• 2022

본 논문에서는 비전공자들을 위한 교양과정으로, 기초 순환신경망 과목 커리큘럼을 설계하는데 필수적으로 요구되는 순환신경망 SW 실습 사례를 개발하였다. 개발된 SW 실습 사례는 순환신경망의 동작원리를 이해시키는 데 초점을 두고, 시각화된 전체 동작 과정을 확인할 수 있도록 스프레드시트를 사용하였다. 개발된 순환신경망 실습 사례는 지도학습 방식의 텍스트완성 훈련데이터 생성, 입력층, 은닉층, 상태층(컨텍스트 노드) 그리고 출력층을 차례대로 구현하고, 텍스트 데이터에 대해 순환신경망의 성능을 테스트하는 것으로 구성되었다. 본 논문에서 개발한 순환신경망 실습사례는 다양한 문자 수를 갖는 단어를 자동 완성한다. 제안한 순환신경망 실습사례를 활용하여, 한글 또는 영어 단어를 구성하는 최대 문자 수를 다양하게 확장하여 자동 완성하는 인공지능 SW 실습 사례를 만들 수 있다. 따라서, 본 순환신경망 기초 실습 사례의 활용도가 높다고 할 수 있다.

• 대한전기학회논문지:전력기술부문A
• /
• 제48권11호
• /
• pp.1464-1471
• /
• 1999

This paper introduces complex-valued competitive learning neural network for spatio-temporal pattern recognition. There have been quite a few neural networks for spatio-temporal pattern recognition. Among them, recurrent neural network, TDNN, and avalanche model are acknowledged as standard neural network paradigms for spatio-temporal pattern recognition. Recurrent neural network has complicated learning rules and does not guarantee convergence to global minima. TDNN requires too many neurons, and can not be regarded to deal with spatio-temporal pattern basically. Grossberg's avalanche model is not able to distinguish long patterns, and has to be indicated which layer is to be used in learning. In order to remedy drawbacks of the above networks, unsupervised competitive learning using complex umber is proposed. Suggested neural network also features simultaneous recognition, time-shift invariant recognition, stable categorizing, and learning rate modulation. The network is evaluated by computer simulation with randomly generated patterns.