• Journal of Korea Multimedia Society
• /
• v.25 no.8
• /
• pp.1203-1211
• /
• 2022

Melanoma is a skin cancer that starts in pigment-producing cells (melanocytes). The death rates of skin cancer like melanoma can be reduced by early detection and diagnosis of diseases. It is common for doctors to spend a lot of time trying to distinguish between skin lesions and healthy cells because of their striking similarities. The detection of melanoma lesions can be made easier for doctors with the help of an automated classification system that uses deep learning. This study presents a new approach for melanoma classification based on an ensemble of deep convolution neural networks and a Log-Gabor filter. First, we create the Log-Gabor representation of the original image. Then, we input the Log-Gabor representation into a new ensemble of deep convolution neural networks. We evaluated the proposed method on the melanoma dataset collected at Yonsei University and Dongsan Clinic. Based on our numerical results, the proposed framework achieves more accuracy than other approaches.

• Journal of the Korean Society for Precision Engineering
• /
• v.13 no.9
• /
• pp.149-155
• /
• 1996

Recent researches in the trouble monitoring system of grinding process have emphasized the use of deep knowledge. Such works include the monitoring and diagnostic systems for cylindrical grinding using sensors on chatter vibration and grinding burn during the process. But, since grinding operations are especially related with a lalrge amount of ambique parameters, it is effectively difficult to detect the grinding troubles occuring during the grinding process. In this paper, monitoring system for grinding utilizes the neural networks based on grinding power signatures. The monitoring system of grinding operations, which makes use of PDP neural networks, is presented. Then, the implementation results by computer simulations and experimental data with respect to chatter vibration and grinding burn are compared.

• Nuclear Engineering and Technology
• /
• v.55 no.6
• /
• pp.2276-2282
• /
• 2023

We deploy artificial neural networks to unfold neutron spectra from measured energy-integrated quantities. These neutron spectra represent an important parameter allowing to compute the absorbed dose and the kerma to serve radiation protection in addition to nuclear safety. The built architectures are inspired from convolutional neural networks. The first architecture is made up of residual transposed convolution's blocks while the second is a modified version of the U-net architecture. A large and balanced dataset is simulated following "realistic" physical constraints to train the architectures in an efficient way. Results show a high accuracy prediction of neutron spectra ranging from thermal up to fast spectrum. The dataset processing, the attention paid to performances' metrics and the hyper-optimization are behind the architectures' robustness.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.28 no.1
• /
• pp.33-58
• /
• 2024

This paper reviews basic concepts for Physics-Informed Neural Networks (PINN) applied to the initial value problems for ordinary differential equations. In particular, using only basic calculus, we derive the error estimates where the error functions (the differences between the true solution and the approximations expressed by neural networks) are dominated by training loss functions. Numerical experiments are conducted to validate our error estimates, visualizing the relationship between the error and the training loss for various first-order differential equations and a second-order linear equation.

• Journal of Institute of Control, Robotics and Systems
• /
• v.9 no.4
• /
• pp.329-339
• /
• 2003

A neural network based adaptive controller design method is proposed for reconfigurable flight control systems in the presence of variations in aerodynamic coefficients or control effectiveness decrease caused by control surface damage. The neural network based adaptive nonlinear controller is developed by making use of the backstepping technique for command following of the angle of attack, sideslip angle, and bank angle. On-line teaming neural networks are implemented to guarantee reconfigurability and robustness to the uncertainties caused by aerodynamic coefficients variations. The main feature of the proposed controller is that the adaptive controller is designed with assumption that not any of the nonlinear functions of the system is known accurately, whereas most of the previous works assume that only some of the nonlinear functions are unknown. Neural networks loam through the weight update rules that are derived from the Lyapunov control theory. The closed-loop stability of the error states is also investigated according to the Lyapunov theory. A nonlinear dynamic model of an F-16 aircraft is used to demonstrate the effectiveness of the proposed control law.

• Computers and Concrete
• /
• v.19 no.3
• /
• pp.293-303
• /
• 2017

Maximum deflection in a beam is a serviceability design criterion and occurs generally at or close to the mid-span. This paper presents a methodology using neural networks for rapid prediction of mid-span deflections in reinforced concrete beams subjected to service load. The closed form expressions are further obtained from the trained neural networks. The closed form expressions take into account cracking in concrete at in-span and at near the interior supports and tension stiffening effect. The expressions predict the inelastic deflections (incorporating the concrete cracking) from the elastic moments and the elastic deflections (neglecting the concrete cracking). Five separate neural networks are trained since these have been postulated to represent all beams having any number of spans. The training, validating, and testing data sets for the neural networks are generated using an analytical-numerical procedure of analysis. The proposed expressions have been verified by comparison with the experimental results reported elsewhere and also by comparison with the finite element method (FEM). The proposed expressions, at minimal input data and minimal computation effort, yield results that are close to FEM results. The expressions can be used in every day design since the errors are found to be small.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.481-481
• /
• 2000

We investigated the structure-activity relationships on use of multi-layer neural networks. The relationships are techniques required in developments of medicines. Since many kinds of observations might be adopted on the techniques, we discussed some points between the observations and the properties of multi-layer neural networks. In the structure-activity relationships, an important property is not that standard deviations are nearly equal to zero for observed physiological activity, but prediction ability for unknown medicines. Since we adopted non-linear approximation, the function to represent the activity can be defined by observations; therefore, we believe that the standard deviations have not significance. The function was examined by "leave-one-out" method, which was originally introduced for the multi-regression analysis. In the linear approximation, the examination is significance, however, we believe that the method is inappropriate in case of nonlinear fitting as neural networks; therefore, we derived a new index fer the relationships from the differential of information propagation in the neural network. By using the index, we discussed physiological activity of an anti-cancer medicine, Mitomycine derivatives. The neuro-computing suggests that there is no direction to extend the anti-cancer activity of Mitomycine, which is close to the trend of anticancer developing.

• Journal of Construction Engineering and Project Management
• /
• v.5 no.1
• /
• pp.26-31
• /
• 2015

Success of the construction companies is based on the successful completion of projects within the agreed cost and time limits. Artificial neural networks (ANN) have recently attracted much attention because of their ability to solve the qualitative and quantitative problems faced in the construction industry. For the estimation of cost and duration different ANN models were developed. The database consists of data collected from completed projects. The same data is normalised and used as inputs and targets for developing ANN models. The models are trained, tested and validated using MATLAB R2013a Software. The results obtained are the ANN predicted outputs which are compared with the actual data, from which deviation is calculated. For this purpose, two successfully completed highway road projects are considered. The Nftool (Neural network fitting tool) and Nntool (Neural network/ Data Manager) approaches are used in this study. Using Nftool with trainlm as training function and Nntool with trainbr as the training function, both the Projects A and B have been carried out. Statistical analysis is carried out for the developed models. The application of neural networks when forming a preliminary estimate, would reduce the time and cost of data processing. It helps the contractor to take the decision much easier.

• Proceedings of the KIEE Conference
• /
• 1998.07b
• /
• pp.552-554
• /
• 1998

In target tracking problems the fixed gain Kalman filter is primarily used to predict a target state vector. This filter, however, has a poor precision for maneuvering targets while it has a good performance for non-maneuvering targets. To overcome the problem this paper proposes the system which estimates the acceleration with neural networks using the input estimation technique. The ability to efficiently fuse information of different forms is one of the major capabilities of trained multi-layer neural networks. The primary motivation for employing neural networks in these applications comes from the efficiency with which more features can be utilized as inputs for estimating target maneuvers. The parallel processing capability of a properly trained neural network can permit fast processing of features to yield correct acceleration estimates. The features used as inputs can be extracted from the combinations of innovation data and heading changes, and for this we set the two dimensional model. The properly trained neural network system outputs the acceleration estimates and compensates for the primary Kalman filter. Finally the proposed system shows the optimum performance.

• Journal of the military operations research society of Korea
• /
• v.34 no.1
• /
• pp.1-11
• /
• 2008

A methodology for the improvement of simulation based training system for the artillery is proposed in this paper. The complex nonlinear relationship inherent among parameters in artillery firing is difficult to model and analyze. By introducing neural network based simulation, accurate representation of artillery firing is made possible. The artillery training system can greatly benefit from the improved prediction. Neural networks learning is conducted using the conjugate gradient algorithm. The evaluation of the proposed methodology is performed through simulation. Prediction errors of both regression analysis model and neural networks model are analyzed. Implementation of neural networks to training system enables more realistic training, improved combat power and reduced budget.