Modified Back-propagation Learning Algorithm Research Articles

A Modified Backpropagation Learning Algorithm With Added Emotional Coefficients

Much of the research work into artificial intelligence (AI) has been focusing on exploring various potential applications of intelligent systems with successful results in most cases. In our attempts to model human intelligence by mimicking the brain structure and function, we overlook an important aspect in human learning and decision making: the emotional factor. While it currently sounds impossible to have "machines with emotions," it is quite conceivable to artificially simulate some emotions in machine learning. This paper presents a modified backpropagation (BP) learning algorithm, namely, the emotional backpropagation (EmBP) learning algorithm. The new algorithm has additional emotional weights that are updated using two additional emotional parameters: anxiety and confidence. The proposed "emotional" neural network will be implemented to a facial recognition problem, and the results will be compared to a similar application using a conventional neural network. Experimental results show that the addition of the two novel emotional parameters improves the performance of the neural network yielding higher recognition rates and faster recognition time.

05/00377 Neuro-PID tracking control of a discharge air temperature system: Zaheer-uddin, M. and Tudoroiu, N. Energy Conversion and Management, 2004, 45, (15–16), 2405–2415

A hybrid neuro-fuzzy control strategy and its corresponding rule generating approach is proposed. According to this approach, the fuzzy control rules can be generated automatically via fuzzy inputs, and then the appropriate control action can be deduced efficiently by a simplified fuzzy inference engine. By combining the use of an incremental PI algorithm and a positional PD algorithm, a PID fuzzy control strategy can be implemented simply from two input variables. It results in the number of control rules being significantly reduced without decreasing the control performance. The control parameters can be self-tuned by introducing a single neuron together with a modified back-propagation learning algorithm. Simulation results show that the proposed fuzzy controller is able to control unknown processes and provide good performance. Compared to traditional self-organising and neural-network-based fuzzy controllers, this method has simpler control algorithms and less computational burden.

DETERMINATION OF ELASTIC CONSTANTS OF ANISOTROPIC LAMINATED PLATES USING ELASTIC WAVES AND A PROGRESSIVE NEURAL NETWORK

In this paper, a procedure is suggested to inversely determine the elastic constants of anisotropic laminated plates using a progressive neural network (NN). The surface displacement responses are used as the inputs for the NN model. The outputs of the NN are the elastic constants of anisotropic laminated plates. The hybrid numerical method (HNM) is used to calculate the displacement responses of laminated plates to an incident wave for given elastic constants. The NN model is trained using the results from the HNM. A modified back-propagation learning algorithm with a dynamically adjusted learning rate and an additional jump factor is developed to tackle the possible saturation of the sigmoid function and to speed up the training process for the NN model. The concept of orthogonal array was adopted to generate the representative combinations of elastic constants, which reduces significantly the number of training data while maintaining its data completeness. Once trained, the NN model can be used for on-line determination of the elastic constants if the dynamic displacement responses on the surface of the laminated plate can be obtained. The determined elastic constants are then used in the HNM to calculate the displacement responses. The NN model would go through a progressive retraining process until the calculated displacement responses using the determined results are sufficiently close to the actual responses. This procedure is examined for an actual glass/epoxy laminated plate. It is found that the present procedure is very robust and efficient for determining the elastic constants of anisotropic laminated plates.

Adaptive multilayer perceptron networks for detection of cracks in anisotropic laminated plates

In this study, an adaptive multiplayer perceptron (MLP) technique is proposed for the detection of cracks in anisotropic laminated plates. The displacement response on the surface of plate, excited by a time-harmonic line load, is used as the input of the MLP. The crack parameters that specify the location and size of the cracks in the anisotropic laminated plates are taken as the output of the MLP. The MLP model is first trained to establish the nonlinear relationship between the scattered surface displacement response and the corresponding location and size of the cracks. The scattered displacement responses required in training samples are calculated from the strip element method (SEM). To facilitate this training process, the correlation analysis for the outputs of neurons in the hidden layers of the MLP model is carried out to optimize the MLP architecture. A modified back-propagation learning algorithm with a dynamically adjusted learning rate and an additional jump factor is developed to speed up the convergence of the MLP model in the training process. The concept of orthogonal array is adopted to generate the representative combinations of the crack parameters, which significantly reduces the number of samples while maintaining the completeness of sample data. The well-trained model is then used to reconstruct the crack parameters by feeding in the measured displacement response on the plate surface. These reconstructed crack parameters are further examined by comparing their resulting displacement response from the SEM forward calculation with the measured displacement response. If the comparison is satisfactory, the reconstructed crack parameters would be considered to be true and the computation ends. Otherwise, the MLP model would go another round of re-training process until the satisfactory reconstruction is obtained. The proposed technique was verified numerically using an anisotropic laminated plate [C0/G+45/G–45] s with four types of horizontal cracks. The verification includes the detection for both the location and the size of cracks using the simulated response data with and without noise.

Dynamic knowledge inference and learning under adaptive fuzzy Petri net framework

Since knowledge in an expert system is vague and modified frequently, expert systems are fuzzy and dynamic. It is very important to design a dynamic knowledge inference framework which is adjustable according to knowledge variation as human cognition and thinking. A generalized fuzzy Petri net model, called adaptive fuzzy Petri net (AFPN), is proposed with this object in mind. AFPN not only has the descriptive advantages of the fuzzy Petri net, it also has learning ability like a neural network. Just as other fuzzy Petri net (FPN) models, AFPN can be used for knowledge representation and reasoning, but AFPN has one important advantage: it is suitable for dynamic knowledge, i.e., the weights of AFPN are adjustable. Based on the AFPN transition firing rule, a modified backpropagation learning algorithm is developed to assure the convergence of the weights.

Structural Damage Detection Using an Iterative Neural Network

A structural damage detection method based on parameter identification using an iterative neural network (NN) technique is proposed in this study. The NN model is first trained off-line using an initial training data set that consists of assumed structural parameters as outputs and their corresponding dynamic characteristics as inputs. The structural parameters are assumed with different levels of reduction to simulate various degrees of structural damage. The concept of orthogonal array is adopted to generate the representative combinations of parameter changes, which can significantly reduce the number of training data while maintaining the data completeness. A modified back-propagation learning algorithm is proposed which can overcome possible saturation of the sigmoid function and speed up the training process. The trained NN model is used to predict the structural parameters by feeding in measured dynamic characteristics. The predicted structural parameters are then used in the FE model to calculate the dynamic characteristics. The NN model would go through a retraining process if the calculated characteristics deviate from the measured ones. The identified structural parameters are then used to infer the location and the extent of structural damages. The proposed method is verified both numerically and experimentally using a clamped-clamped T beam. The results indicate that the current approach can identify both the location and the extent of damages in the beam.

PARALLEL NEURAL LEARNING FOR CONTROL PROBLEMS ON A BUS-BASED ARCHITECTURE

In [6], we distributed training instances over a single-channel broadcast communication model to speed up execution of the back-propagation learning algorithm for classification problems. In this paper, we extend this concept to control problems, where the output is not necessarily 0 or 1, but ranges over an interval. We first propose a modified back-propagation learning algorithm that incrementally decreases the error threshold by half in order to process training instances with large weight changes as quickly as possible. This modified back-propagation learning algorithm is then parallelized using the single-channel broadcast communication model to n processors, where n is the number of training instances. Finally, the parallel back-propagation learning algorithm is modified for execution on a bounded number of processors to cope with real-world conditions.

A hybrid neuro-fuzzy PID controller

A hybrid neuro-fuzzy control strategy and its corresponding rule generating approach is proposed. According to this approach, the fuzzy control rules can be generated automatically via fuzzy inputs, and then the appropriate control action can be deduced efficiently by a simplified fuzzy inference engine. By combining the use of an incremental PI algorithm and a positional PD algorithm, a PID fuzzy control strategy can be implemented simply from two input variables. It results in the number of control rules being significantly reduced without decreasing the control performance. The control parameters can be self-tuned by introducing a single neuron together with a modified back-propagation learning algorithm. Simulation results show that the proposed fuzzy controller is able to control unknown processes and provide good performance. Compared to traditional self-organising and neural-network-based fuzzy controllers, this method has simpler control algorithms and less computational burden.

Modeling nuclear reactor core dynamics with recurrent neural networks

A recurrent multilayer perception (RMLP) model is designed and developed for simulation of core neutronic phenomena in a nuclear power plant, which constitute a non-linear, complex dynamic system characterized by a large number of state variables. Training and testing data are generated by REMARK, a first principles neutronic core model [16]. A modified backpropagation learning algorithm with an adaptive steepness factor is employed to speed up the training of the RMLP. The test results presented exhibit the capability of the recurrent neural network model to capture the complex dynamics of the system, yielding accurate predictions of the system response. The performance of the network is also demonstrated for interpolation, extrapolation, fault tolerance due to incomplete data, and for operation in the presence of noise.

Intelligent stock trading system with price trend prediction and reversal recognition using dual-module neural networks

This article presents an intelligent stock trading system that can generate timely stock trading suggestions according to the prediction of short-term trends of price movement using dual-module neural networks(dual net). Retrospective technical indicators extracted from raw price and volume time series data gathered from the market are used as independent variables for neural modeling. Both neural network modules of thedual net learn the correlation between the trends of price movement and the retrospective technical indicators by use of a modified back-propagation learning algorithm. Reinforcing the temporary correlation between the neural weights and the training patterns, dual modules of neural networks are respectively trained on a short-term and a long-term moving-window of training patterns. An adaptive reversal recognition mechanism that can self-tune thresholds for identification of the timing for buying or selling stocks has also been developed in our system. It is shown that the proposeddual net architecture generalizes better than one single-module neural network. According to the features of acceptable rate of returns and consistent quality of trading suggestions shown in the performance evaluation, an intelligent stock trading system with price trend prediction and reversal recognition can be realized using the proposed dual-module neural networks.