• The Korean Journal of Applied Statistics
• /
• v.26 no.5
• /
• pp.847-856
• /
• 2013

We propose a modified learning process for generalized neural network using a learning algorithm by Liu et al. (2001). We consider the effect of initial weights, training results and learning errors using a modified learning process. We employ an incremental training procedure where training patterns are learned systematically. Our algorithm starts with a single training pattern and a single hidden layer neuron. During the course of neural network training, we try to escape from the local minimum by using a weight scaling technique. We allow the network to grow by adding a hidden layer neuron only after several consecutive failed attempts to escape from a local minimum. Our optimization procedure tends to make the network reach the error tolerance with no or little training after the addition of a hidden layer neuron. Simulation results with suitable initial weights indicate that the present constructive algorithm can obtain neural networks very close to minimal structures and that convergence to a solution in neural network training can be guaranteed. We tested these algorithms extensively with small training sets.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.10
• /
• pp.4117-4135
• /
• 2020

Due to the angle of view, illumination and scene diversity, real-time dynamic face detection and recognition is no small difficulty in those unrestricted environments. In this study, we used the intrinsic correlation between detection and calibration, using a multi-task cascaded convolutional neural network(MTCNN) to improve the efficiency of face recognition, and the output of each core network is mapped in parallel to a compact Euclidean space, where distance represents the similarity of facial features, so that the target face can be identified as quickly as possible, without waiting for all network iteration calculations to complete the recognition results. And after the angle of the target face and the illumination change, the correlation between the recognition results can be well obtained. In the actual application scenario, we use a multi-camera real-time monitoring system to perform face matching and recognition using successive frames acquired from different angles. The effectiveness of the method was verified by several real-time monitoring experiments, and good results were obtained.

• Structural Engineering and Mechanics
• /
• v.12 no.5
• /
• pp.527-540
• /
• 2001

A new sort of learning algorithm named whole learning algorithm is proposed to simulate the nonlinear and dynamic behavior of RC members for the estimation of structural integrity. A mathematical technique to solve the multi-objective optimization problem is applied for the learning of the feedforward neural network, which is formulated so as to minimize the Euclidean norm of the error vector defined as the difference between the outputs and the target values for all the learning data sets. The change of the outputs is approximated in the first-order with respect to the amount of weight modification of the network. The governing equation for weight modification to make the error vector null is constituted with the consideration of the approximated outputs for all the learning data sets. The solution is neatly determined by means of the Moore-Penrose generalized inverse after summarization of the governing equation into the linear simultaneous equations with a rectangular matrix of coefficients. The learning efficiency of the proposed algorithm from the viewpoint of computational cost is verified in three types of problems to learn the truth table for exclusive or, the stress-strain relationship described by the Ramberg-Osgood model and the nonlinear and dynamic behavior of RC members observed under an earthquake.

• International Journal of Fluid Machinery and Systems
• /
• v.3 no.1
• /
• pp.29-38
• /
• 2010

This paper presents a procedure for the design optimization of a centrifugal compressor. The centrifugal compressor consists of a centrifugal impeller, vaneless diffuser and volute. And, optimization techniques based on the radial basis neural network method are used to optimize the impeller of a centrifugal compressor. The Latin-hypercube sampling of design-of-experiments is used to generate the thirty design points within design spaces. Three-dimensional Reynolds-averaged Navier-Stokes equations with the shear stress transport turbulence model are discretized by using finite volume approximations and solved on hexahedral grids to evaluate the objective function of the total-to-total pressure ratio. Four variables defining the impeller hub and shroud contours are selected as design variables in this optimization. The results of optimization show that the total-to-total pressure ratio of the optimized shape at the design flow coefficient is enhanced by 2.46% and the total-to-total pressure ratios at the off-design points are also improved significantly by the design optimization.

• Proceedings of the IEEK Conference
• /
• 1998.06a
• /
• pp.130-133
• /
• 1998

A imput buffer typed ATM switch and an appropriate cell-scheduling algorithm are necessary for avoiding output blocking and internal blocking respectively. The algorithm determining a set of non-blocking data cells from the queues can greatly affect on the switch's throughput as well as the behavior of the queues. In this paper bit pattern optimization combined with the Token method in presented in order to improve the performance of ATM switch. The digital Hopfield neural cell scheduler is designed and used for the maximum numbers of cells in real-time

• Journal of the Korean Institute of Intelligent Systems
• /
• v.7 no.2
• /
• pp.87-95
• /
• 1997

This paper presents an optimal design procedure to realize an BSB neural networks by means of the parametrization of solution space and optimization of parameters using evaluation program. In particular, the performance index based on DOA analysis may make an associative memory implementation reach on the level of practical success.

• Proceedings of the KIEE Conference
• /
• 1990.07a
• /
• pp.492-495
• /
• 1990

This paper introduces a path planning technique for a mobile robot in the presence of obstacles. In the technique, workspace is described by regional graph and represented obstacles by the three-layer neural network. And performance cost is defined under consideration both the traveling distance and the safety of a mobile robot. Then a collision-free path is obtained using the neural optimization technique.

• The KIPS Transactions:PartB
• /
• v.18B no.5
• /
• pp.271-278
• /
• 2011

Storm and flood such as torrential rains and major typhoons has often caused damages on a large scale in Korea and damages from storm and flood have been increasing by climate change and warming. Therefore, it is an essential work to maneuver preemptively against risks and damages from storm and flood by predicting the possibility and scale of the disaster. Generally the research on numerical model based on statistical methods, the KDF model of TCDIS developed by NIDP, for analyzing and predicting disaster risks and damages has been mainstreamed. In this paper, we introduced the model for prediction of damage cost from storm and flood by the neural network algorithm which outstandingly implements the pattern recognition. Also, we compared the performance of the neural network model with that of KDF model of TCDIS. We come to the conclusion that the robustness and accuracy of prediction of damage cost on TCDIS will increase by adapting the neural network model rather than the KDF model.

• Proceedings of the Korean Institute Of Construction Engineering and Management
• /
• 2004.11a
• /
• pp.317-322
• /
• 2004

As a construction project in urban area tends to be high-rise and huge, the importance of the project's underground work, in terms of the cost and the schedule, is gradually increasing. It's extremely significant to choose a proper filing method, at the stage of underground work. However, in piling work many change orders have been occurred since a piling method is experientially selected based on uncertain information and many earth factors to consider. It has effects on the cost and the schedule of the project. In this study, we have suggested a decision model for piling method that can be used to determine and verify the suitable piling method in design and pre-construction phase of a project. Based on historical data, a neural network model has already proven to be efficient. The tests of the model for selecting a suitable piling method have progressed exactly with the data of 150 piling works which were done room 2000 to 2004 in Korea. The optimization or the developed neural network model has progressed with the data for teaming. The validity of the neural network model has been verified.

• The Journal of the Acoustical Society of Korea
• /
• v.42 no.4
• /
• pp.305-312
• /
• 2023

Physics-Informed Neural Network (PINN) is used to invert bubble size distributions from attenuation losses. By considering a linear system for the bubble population inversion, Adaptive Learned Iterative Shrinkage Thresholding Algorithm (Ada-LISTA), which has been solved linear systems in image processing, is used as a neural network architecture in PINN. Furthermore, a regularization based on the linear system is added to a loss function of PINN and it makes a PINN have better generalization by a solution satisfying the bubble physics. To evaluate an uncertainty of bubble estimation, deep ensemble is adopted. 20 Ada-LISTAs with different initial values are trained using the same training dataset. During test with attenuation losses different from those in the training dataset, the bubble size distribution and corresponding uncertainty are indicated by average and variance of 20 estimations, respectively. Deep ensemble Ada-LISTA demonstrate superior performance in inverting bubble size distributions than the conventional convex optimization solver of CVX.