• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.1
• /
• pp.1-10
• /
• 2022

Deep neural networks are widely used to solve various problems. In a fully connected neural network, the nonlinear activation function is a function that nonlinearly transforms the input value and outputs it. The nonlinear activation function plays an important role in solving the nonlinear problem, and various nonlinear activation functions have been studied. In this study, we propose a combined parametric activation function that can improve the performance of a fully connected neural network. Combined parametric activation functions can be created by simply adding parametric activation functions. The parametric activation function is a function that can be optimized in the direction of minimizing the loss function by applying a parameter that converts the scale and location of the activation function according to the input data. By combining the parametric activation functions, more diverse nonlinear intervals can be created, and the parameters of the parametric activation functions can be optimized in the direction of minimizing the loss function. The performance of the combined parametric activation function was tested through the MNIST classification problem and the Fashion MNIST classification problem, and as a result, it was confirmed that it has better performance than the existing nonlinear activation function and parametric activation function.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.7
• /
• pp.1773-1780
• /
• 1996

KBANN (knowledge-based artificial neural network) combining the symbolic approach and the numerical approach has been shown to be more effective than other machine learning models. However KBANN doesn't have the theory refinement ability because the topology of network can't be altered dynamically. Although TopGen was proposed to extend the ability of KABNN in this respect, it also had some defects due to the link-ing of hidden nodes to input nodes and the use of beam search. The algorithm which could solve this TopGen's defects, by adding the hidden nodes linked to next layer nodes and using hill-climbing search with backtracking, is designed.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• Proceedings of the Materials Research Society of Korea Conference
• /
• 1993.05a
• /
• pp.61-61
• /
• 1993

20세기 막바지에 들어 서면서 21세기를 바라보는 정보화사회는 어떠한 과학기술의발전을 필요로 하는 것일까? $10^{12}$ 집적도, $10^{-8}$cm 원자공간, $10^{-12}$-$10^{-15}$ 초등 시간대역에 대한 도전과[ 테라바이트 메모리], [테라바이트 컴퓨터], [테라비트 광통신] 등을 앞세우면서 고속화, 대용량화, 초미세화, 다기능화, 고기능화, 지능화를 지향하는 미래 정보통신 기술 실용화를 위하여서는 어떠한 성질의 신소재들이 개발되어야 할 것인가? [20세기 전자시대] 의 대표적 정보운반자인 [전자] 에 대한 연구는 21세기에는 어떻게 전개될 것인가? 새로운 정보운반자로서 부상하고 있는 [광자],[뉴런],[생체분자]등에 대한 연구는 어떠한 방향으로 전개될 것인가? 신개념의 정보통신 기술을 구체적으로 실용화 하기 위하여 연구되어야 할 신소재는 어떻게 전개되어야 할 것인가\ulcorner 초미세구조, 양자효과, 비선형효과, 원자가공, 원자조작, 인공신소재, 초격자 지능신소재, 초전도 유기물, 분자, 광논리, 광신경망, 생체노리, 생체컴퓨터등 신개념의 창출로부터 비롯해서 의료, 복지, 장애. 기후. 환경. 지각.해양. 항공. 우주에 이르는 다차원적 통신과 지능형 정보를 가능케 하는 신소재 연구의 조건들과 그에 따른 도전을 전망해 본다.에 따른 도전을 전망해 본다.

• Journal of the Korean Society for Precision Engineering
• /
• v.22 no.4
• /
• pp.37-43
• /
• 2005

Real time monitoring of exit burr formation is critical in manufacturing automation. In this paper, acoustic emission (AE) was used to detect the burr formation during drilling. By using wavelet transform (WT), AE data were compressed without unnecessary details. Then the transformed data were used as selected features (inputs) of a back-propagation artificial neural net (ANN). In order to validate the in process AE monitoring system, both WT-based ANN and cutting condition (cutting speed, feed, drill diameter, etc.) based ANN outputs were compared with experimental data.

• Journal of the Korean Society for Precision Engineering
• /
• v.20 no.11
• /
• pp.71-78
• /
• 2003

The exit burrs in the micro-drilling of precision miniature holes are of interest, especially for ductile materials. As burrs from this process can be difficult to remove, it is important to acquire the way of predicting burr types as well as optimal cutting conditions which minimize the burrs. In this paper, an artificial neural network was used for the prediction of burr formation in micro-drilling. First, the influence of cutting conditions including cutting speed, feed and drill diameter on the exit burr characteristics, such as burr size and type, were observed and analyzed. Then. the burr types were classified by using the influential experimental data as input parameters to the neural nets.

• Journal of Intelligence and Information Systems
• /
• v.23 no.1
• /
• pp.109-126
• /
• 2017

In the various kinds of intelligent services, many studies for detecting emotion are in progress. Particularly, studies on emotion recognition at the particular time have been conducted in order to provide personalized experiences to the audience in the field of exhibition though facial expressions change as time passes. So, the aim of this paper is to build a model to predict the audience's emotion from the changes of facial expressions while watching an exhibit. The proposed model is based on both sequential neural network and the Valence-Arousal model. To validate the usefulness of the proposed model, we performed an experiment to compare the proposed model with the standard neural-network-based model to compare their performance. The results confirmed that the proposed model considering time sequence had better prediction accuracy.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.33 no.5
• /
• pp.514-520
• /
• 2009

In configuring an automated polishing system, a monitoring scheme to estimate the surface roughness is necessary. In this study, a precision polishing process, magnetic abrasive finishing (MAF), along with an in-process monitoring setup was investigated. A magnetic tooling is connected to a CNC machining to polish the surface of stavax(S136) die steel workpieces. During finishing experiments, both AE signals and force signals were sampled and analysed. The finishing results show that MAF has nano scale finishing capability (upto 8nm in surface roughness) and the sensor signals have strong correlations with the parameters such as gap between the tool and workpiece, feed rate and abrasive size. In addition, the signals were utilized as the input parameters of artificial neural networks to predict generated surface roughness. Among the three networks constructed -AE rms input, force input, AE+force input- the ANN with sensor fusion (AE+force) produced most stable results. From above, it has been shown that the proposed sensor fusion scheme is appropriate for the monitoring and prediction of the nano scale precision finishing process.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.26 no.4
• /
• pp.382-391
• /
• 2020

In this study, we used artificial intelligence algorithms for the prediction of dissolved oxygen in Jindong Bay. To determine missing values in the observational data, we used the Bidirectional Recurrent Imputation for Time Series (BRITS) deep learning algorithm, Auto-Regressive Integrated Moving Average (ARIMA), a widely used time series analysis method, and the Long Short-Term Memory (LSTM) deep learning method were used to predict the dissolved oxygen. We also compared accuracy of ARIMA and LSTM. The missing values were determined with high accuracy by BRITS in the surface layer; however, the accuracy was low in the lower layers. The accuracy of BRITS was unstable due to the experimental conditions in the middle layer. In the middle and bottom layers, the LSTM model showed higher accuracy than the ARIMA model, whereas the ARIMA model showed superior performance in the surface layer.

• Journal of Intelligence and Information Systems
• /
• v.14 no.3
• /
• pp.25-39
• /
• 2008

The objective of this study is to forecast the operational continuous ability using Artificial Neural Networks in battalion defensive operation for the commander decision making support. The forecasting of the combat result is one of the most complex issue in military science. However, it is difficult to formulate a mathematical model to evaluate the combat power of a battalion in defensive operation since there are so many parameters and high temporal and spatial variability among variables. So in this study, we used company combat power level data in Battalion Command in Battle Training as input data and used Feed-Forward Multilayer Perceptrons(MLP) and General Regression Neural Network (GRNN) to evaluate operational continuous ability. The results show 82.62%, 85.48% of forecasting ability in spite of non-linear interactions among variables. We think that GRNN is a suitable technique for real-time commander's decision making and evaluation of the commitment priority of troops in reserve.