• Geomechanics and Engineering
• /
• v.37 no.3
• /
• pp.263-277
• /
• 2024

Artificial Neural Networks (ANNs) are artificial learning algorithms that provide successful results in solving many machine learning problems such as classification, prediction, object detection, object segmentation, image and video classification. There is an increasing number of studies that use ANNs as a prediction tool in soil classification. The aim of this research was to understand the role of hyperparameter optimization in enhancing the accuracy of ANNs for soil type classification. The research results has shown that the hyperparameter optimization and hyperparamter optimized ANNs can be utilized as an efficient mechanism for increasing the estimation accuracy for this problem. It is observed that the developed hyperparameter tool (HyperNetExplorer) that is utilizing the Covariance Matrix Adaptation Evolution Strategy (CMAES), Genetic Algorithm (GA) and Jaya Algorithm (JA) optimization techniques can be successfully used for the discovery of hyperparameter optimized ANNs, which can accomplish soil classification with 100% accuracy.

• The Journal of the Acoustical Society of Korea
• /
• v.38 no.6
• /
• pp.703-709
• /
• 2019

This paper proposes speech recognition systems employing Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM) structures combined with Hidden Markov Moldel (HMM) to effectively recognize the speech of VeloPharyngeal Insufficiency (VPI) patients, and compares the recognition performance of the systems to the Gaussian Mixture Model (GMM-HMM) and fully-connected Deep Neural Network (DNNHMM) based speech recognition systems. In this paper, the initial model is trained using normal speakers' speech and simulated VPI speech is used for generating a prior model for speaker adaptation. For VPI speaker adaptation, selected layers are trained in the CNN-HMM based model, and dropout regulatory technique is applied in the LSTM-HMM based model, showing 3.68 % improvement in recognition accuracy. The experimental results demonstrate that the proposed LSTM-HMM-based speech recognition system is effective for VPI speech with small-sized speech data, compared to conventional GMM-HMM and fully-connected DNN-HMM system.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.55 no.4
• /
• pp.154-157
• /
• 2006

This paper is proposed hybrid artificial intelligent(HAI) controller for high performance of induction motor drive. The design..of this algorithm based on fuzzy-neural network(FNN) controller that is implemented using fuzzy control and neural network. This controller uses fuzzy rule as training patterns of a neural network. Also, this controller uses the back-propagation method to adjust the weights between the neurons of neural network in order to minimize the error between the command output and actual output. A model reference adaptive scheme is proposed in which the adaptation mechanism is executed by fuzzy logic based on the error and change of error measured between the motor speed and output of a reference model. The control performance of the adaptive FNN controller is evaluated by analysis for various operating conditions. The results of experiment prove that the proposed control system has strong high performance and robustness to parameter variation, and steady-state accuracy and transient response.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2000.11a
• /
• pp.376-379
• /
• 2000

In this paper, this system makes use of the analog infrered rays sensor and converts the feature of fish analog signal when sensor is operating with CPU(80C196KC). Then, After signal processing, this feature is classified a special feature and a outline of fish by using the neural network, one of the artificial intelligence scheme. This neural network classifies fish pattern of very simple and short calculation. This has linear activation function and the error backpropagation is used as a learning algorithm. And the neural network is learned in off-line process. Because an adaptation period of neural network is too long time when random initial weights are used, off-line learning is induced to decrease the progress time. We confirmed this method has better performance than somewhat outdated machines.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.158-163
• /
• 1998

This paper is our first attempt to construct a information processing system such as the living creatures' brain based on artificial life technique. In this paper, we propose a method of constructing neural networks using bio-inspired emergent and evolutionary concept, Ontogeny of living things is realized by cellular automata model and Phylogeny that is living things adaptation ability themselves to given environment, are realized by evolutionary algorithms. Proposing evolving cellular automata neural systems are calledin a word ECANS. A basic component of ECANS is 'cell' which is modeled on chaotic neuron with complex characteristics, In our system, the states of cell are classified into eight by method of connection neighborhood cells. When a problem is given, ECANS adapt itself to the problem by evolutionary method. For fixed cells transition rule, the structure of neural network is adapted by change of initial cell' arrangement. This initial cell is to become a network b developmental process. The effectiveness and the capability of proposed scheme are verified by applying it to pattern classification and robot control problem.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.41 no.3
• /
• pp.39-46
• /
• 2004

This paper is proposed adaptive fuzzy-neural network(FNN) controller for the speed control of interior permanent magnet synchronous motor(IPMSM) drive. The design of this algorithm based on FNN controller that is implemented by using fuzzy control and neural network. This controller uses fuzzy rule as training patterns of a neural network. Also, this controller uses the back-propagation method to adjust the weights among the neurons of neural network in order to minimize the error between the command output and actual output. A model reference adaptive scheme is proposed in which the adaptation mechanism is executed by fuzzy logic based on the error and change of error measured between the motor speed and output of a reference model. The control performance of the adaptive FNN controller is evaluated by analysis for various operating conditions. The results of analysis prove that the proposed control system has strongly high performance and robustness in parameter variation, steady-state accuracy and transient response.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.8 no.6
• /
• pp.1135-1141
• /
• 2004

In this study, this system makes use of the analog infrared rays sensor and converts the feature of fish outline when sensor is operating with CPU(80C196KC). Then, after signal processing, this feature is classified a special feature and a outline of fish by using the neural network, one of the artificial intelligence scheme. This neural network classifies fish pattern of very simple and short calculation. This has linear activation function and the error back propagation is used as a teaming algorithm. And the neural network is learned in off-line process. Because an adaptation period of neural network is too long when random initial weights are used, off-line teaming is induced to decrease the progress time.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.7 no.5
• /
• pp.34-45
• /
• 1997

In this paper, we present hardware implemcntation of self-organizing feature map (SOFM) neural networkwith constant adaptation gain and binary reinforcement function on FPGA. Whereas a tnme-varyingadaptation gain is used in the conventional SOFM, the proposed SOFM has a time-invariant adaptationgain and adds a binary reinforcement function in order to compensate for the lowered abilityof SOFM due to the constant adaptation gain. Since the proposed algorithm has no multiplication operation.it is much easier to implement than the original SOFM. Since a unit neuron is composed of 1adde $r_tracter and 2 adders, its structure is simple, and thus the number of neurons fabricated onFPGA is expected to he large. In addition, a few control signal: ;:rp sufficient for controlling !he neurons.Experimental results show that each componeni ot thi inipiemented neural network operates correctlyand the whole system also works well.stem also works well.

• The Journal of Korean Association of Computer Education
• /
• v.16 no.2
• /
• pp.39-48
• /
• 2013

This study investigates the predictive model of the adolescents' computer entertainment behavior using neural network with the KYPS data (3449 in the junior high school; 1725 boys and 1724 girls). This study compares the results of neural network(model 1) to the logistic regression model and neural network(model 2) with the exact same variables used in logistic regression. The results reveal that the prediction of neural network model 1 is the highest among three models and with gender, computer use time, family income, the number of close friends, the number of misdeed friends, individual study time, self-control, private education time, leisure time, self-belief, stress, adaptation to school, and study related worries, the neural network model 1 predicts the computer entertainment behavior more efficiently. These results suggest that the neural network could be used for diagnosing and adjusting the adolescents' computer entertainment behavior.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.24 no.4
• /
• pp.382-391
• /
• 2021

Unsupervised domain adaptation often gives impressive solutions to handle domain shift of data. Most of current approaches assume that unlabeled target data to train is abundant. This assumption is not always true in practices. To tackle this issue, we propose a general solution to solve the domain gap minimization problem without any target data. Our method consists of two regularization steps. The first step is a pixel regularization by arbitrary style transfer. Recently, some methods bring style transfer algorithms to domain adaptation and domain generalization process. They use style transfer algorithms to remove texture bias in source domain data. We also use style transfer algorithms for removing texture bias, but our method depends on neither domain adaptation nor domain generalization paradigm. The second regularization step is a feature regularization by feature alignment. Adding a feature alignment loss term to the model loss, the model learns domain invariant representation more efficiently. We evaluate our regularization methods from several experiments both on small dataset and large dataset. From the experiments, we show that our model can learn domain invariant representation as much as unsupervised domain adaptation methods.