Shang-Jen Chuang

Quantum NN vs. NN in signal recognition

In this paper, the signal recognition by using quantum neural network (QNN) is studied and simulated. The signals with fuzziness distributed in the boundary of two different types of signals could be effectively recognized due to the structure of QNNs hidden units. To demonstrate the capability of QNN in recognition, the signals in a two-dimension (NC2) non-convex system is simulated. All the experiments are also performed by using the traditional neural network (NN) for a comparison.

Fast learning neural network with modified neurons

In this paper, a neural model with modified neurons is developed. Compare with traditional neural network, such a neural model not only has a fast learning, but also can significantly improve the accuracy while it is used in real applications. To demonstrate the learning efficiency of neural model we developed, a nonlinear system identification problem is studied and simulated. All simulations are performed by using constant learning rates (0.1/spl sim/0.9). From the simulation results, the neural model we developed obviously has excellent performances as desired.

The model reference control by adaptive PID-like fuzzy-neural controller

In this paper, an adaptive PID-Like fuzzy-neural controller is proposed and applied to nonlinear model reference control system. For enhancing the flexibility and control capability of the controller we developed, three parameters, including the immediate system error (e(k)), error change (e/spl dot/(k)) and the change of error change (e/spl uml/(k)), are used as the reference inputs for fuzzy-neural tuning mechanism. To demonstrate the superiority of the controller we designed, two types of model reference control systems are studied and simulated. For a comparison, same experiments are also performed by using conventional fuzzy controller with same fuzzy mechanism.

The Transmittance Estimation of Touch Panel Decoration Film by Neural Network

In this paper, a transmittance estimator of touch panel decoration film by using neural network is presented. In the evaporation process, the coating material and the related control parameters are all important influencing factors in obtaining the desired transmittance. The relationship among the transmittance and these factors are very complex and nonlinear. It’s very hard to use the certain mathematical model to describe such relationship. In this research, the neural network was employed to catch the relationship among transmittance and its possible influencing factors. In other words, an efficiently and precisely automatic evaporation parameters decision system for touch panel decoration film is expected to be developed. Through the estimation system developed, the quality of transmittance of touch panel film could meet the customer’s requirement.

Power Signal Forecasting by Neural Model with Different Layer Structures

In this paper, the non-stationary power load forecasting by using neural model with different layer structures is presented. In the neural forecasting model we developed, the neuron types used in different layers are different. Each layer is composed of the same kind of neurons. A reliable and accurate neural forecasting model for the non-stationary power loads is trying to be found in this study. To demonstrate the superiority of the model we created, all simulations are executed by using the conventional neural model with same neurons as a comparison. From the results shown, it is clearly found that the neural model we constructed do have better nonlinear mapping and forecasting capabilities in comparison with the conventional neural model

A New Technique for Searching the Global Minimum of Supervised Neural Network

This paper presents a technique in how to searching the global minimum for the supervised neural network training. This technique is developed based on the idea of nearly equivalent model. To demonstrate the new technique proposed, two signal processing studies, including signal recognition and signal modeling were simulated. For a comparison, the same simulations were also performed by using the neural network with the standard steepest descent error back-propagation (BP) algorithm. From the simulation results shown, the technique we proposed not only can evidence whether the neural network is in the local training or not, but also can show that the “best” performance of the neural network should have.

The chromatic aberration estimation of TP film by using quasi-Newton neural networks

This paper presents the chromatic aberration estimations of touch panel (TP) film by using quasi-Newton neural networks. The data of TP film with one layer coating was studied and simulated. Through the training of neural network, the complex relationship between the chromatic aberration, i.e., L.A.B. values, and the relative parameters of TP decoration film can be obtained. From the simulation results shown, the estimation of chromatic aberration of TP film is quite accurate and promising. That means an artificial intelligent (AI) estimator for the physical properties of TP film is possibly developed. Based on this AI estimator, the relative control parameters of evaporation process could be set in advance such that the quality of TP could meet the customers request.

Quality Identification of the Riveting Process by QNN Model

In this paper, an automatic quality inspection system for the riveting process by using quantum neural network (QNN) was proposed. This inspection system not only can monitor the real time riveting process, but also can give the assistance on the riveting quality verification. For demonstrating the superiority of the inspection system we developed, the data provided by the experiment did by Chinese Air Force Institute of Technology was simulated. The method of riveting quality index (RQI) was also performed as a comparison.

The Prediction of Ink Thickness of Touch Panel by Neural Network

The thickness of printing ink on touch panel is an important work, which is related with the quality of panel in real use. Therefore, how to make a well control in the film printing of touch panel is very important work in the manufacturing process of touch panel. In this paper, the prediction of ink thickness of touch panel based on neural technique is proposed. This intelligent predictor is expected to provide the accurate information about the control parameters so that the technician could make a good setting work before the panel is on the real-line manufacturing operation. Without enough experiences, the junior technician still can make a good decision in the setting of control parameters for ink printing based on this predictor. Thus, this intelligent prediction system not only can help technicians and greatly improve their working efficiency, but also can save the cost of production.

Power Signal Predictions by Using Modified Neural Network

In this paper, the non-stationary power signal prediction by using modified neural network (NN) is presented. Due to the special structure of neuron used, the NN model proposed not only has a fast learning speed, the prediction accuracy in power signal is much better as compared with conventional NN models. To demonstrate the superiority of modified NN model we proposed, all simulations are executed by using three different NN models as a comparison.