• 정보보호학회논문지
• /
• 제33권5호
• /
• pp.761-773
• /
• 2023

본 논문에서는 딥러닝 분야에서 사용되는 신경망 모델, 그중에서도 다중 계층 퍼셉트론 모델에 사용되는 지수함수 기반의 활성화 함수를 근사 함수로 대체하고, 근사 함수에 마스킹을 적용함으로써 신경망 모델의 추론 과정의 전력 분석 저항성을 높이는 방법을 제안한다. 이미 학습된 값을 사용하여 연산하는 인공 신경망의 추론 과정은 그 특성상 가중치나 편향 등의 내부 정보가 부채널 공격에 노출될 위험성이 있다. 다만 신경망 모델의 활성화 함수 계층에서는 매우 다양한 함수를 사용하고, 특히 지수함수 기반의 활성화 함수에는 마스킹 기법 등 통상적인 부채널 대응기법을 적용하기가 어렵다. 따라서 본 연구에서는 지수함수 기반의 활성화 함수를 단순한 형태로 근사하여도 모델의 치명적인 성능 저하가 일어나지 않음을 보이고, 근사 함수에 마스킹을 적용함으로써 전력 분석으로부터 안전한 순방향 신경망 모델을 제안하고자 한다.

• 산업진흥연구
• /
• 제6권1호
• /
• pp.1-7
• /
• 2021

본 연구는 다층 퍼셉트론과 지도형 학습알고리즘에 대해 알아보았고, 아울러 neuralnet이라는 패키지를 사용하여 공변수들과 반응변수 간의 함수적 관계를 어떻게 모델링하는지 살펴보았다. 본 연구에서 적용된 알고리즘은 반응변수 값의 실제치와 예측치 간의 비교에 근거한 오차함수의 최소화를 위한 모수인 가중치들의 계속적인 조정을 특징으로 한다. 본 연구에서 설명하는 neuralnet 패키지는 활성화함수와 오차함수를 주어진 상황에 맞게 적절히 선택하고 나머지 매개변수들은 기본값으로 둘 수 있다. 본 연구에서 살펴본 불임 데이터에 대해 neuralnet 패키지를 활용한 결과 4개의 독립변수 중에서 age는 불임에 영향력이 거의 없음을 파악할 수 있었다. 아울러 신경망의 가중치는 -751.6부터 7.25에 이르기까지 다양한 값을 취하며, 첫 번째 은닉층의 절편은 -92.6과 7.25이며, 첫 번째 은닉뉴런으로 가는 공변수 age, parity, induced, spontaneous에 대한 가중치는 각각 3.17, -5.20, -36.82, -751.6임을 파악했다.

• 한국음향학회지
• /
• 제13권6호
• /
• pp.5-14
• /
• 1994

이 논문은 오차역전달(error back-propagation) 알고리듬을 갖는 다층구조 퍼셉트런(Multi-Layered Perceptron)을 사용하여 우리말 단어음성을 화자종속으로 기계 인식하는 실험에 관한 연구 결과다. 대상단어는 시외 자동전화 지역번호표에서 임의로 선택한 50개 지역명이며, 이 중 43개는 2음절로 구성되어있고 나머지 7개는 3음절이다. 단어를 음소나 음절별로 분리(segmentation)하지 않고, 단어의 각 부분에서 골고루 추출된 특징성분을 신경망에 입력하는 방법을 사용했다. 그렇게 함으로써 발음지속시간에 관계없는 결과를 얻을 수 있으며, 이 때 사용된 특징 성분은 선형예측분석으로 구해진 PARCOR계수다. 전체학습과 구분학습의 비교, 프레임 갯수와 PARCOR차수에 대한 인식률의 의존도, 중간층 뉴런의 갯수에 대한 인식률의 변동, 그리고 출력층 뉴런의 구성 방법에 따른 비교 등 4가지 실험을 통하여 가장 최량의 조건을 찾아보고자 하였다. 이 연구를 발전시킨다면 실시간의 화자독립 소규모어휘 음성인식이 가능해질 것으로 보인다.

• 제어로봇시스템학회논문지
• /
• 제4권4호
• /
• pp.487-493
• /
• 1998

Camera calibration is the process of determining the coordinate relationship between a camera image and its real world space. Accurate calibration of a camera is necessary for the applications that involve quantitative measurement of camera images. However, if the camera plane is parallel or near parallel to the calibration board on which 2 dimensional objects are defined(this is called "ill-conditioned"), existing solution procedures are not well applied. In this paper, we propose a neural network-based approach to camera calibration for 2D images formed by a mono-camera or a pair of cameras. Multi-layer perceptrons are developed to transform the coordinates of each image point to the world coordinates. The validity of the approach is tested with data points which cover the whole 2D space concerned. Experimental results for both mono-camera and stereo-camera cases indicate that the proposed approach is comparable to Tsai's method［8］. Especially for the stereo camera case, the approach works better than the Tsai's method as the angle between the camera optical axis and the Z-axis increases. Therefore, we believe the approach could be an alternative solution procedure for the ill -conditioned camera calibration.libration.

• Computers and Concrete
• /
• 제10권6호
• /
• pp.557-573
• /
• 2012

This paper discusses the combined application of two different techniques, Neural Networks (NN) and Principal Component Analysis (PCA), for improved prediction of concrete properties. The combination of these approaches allowed the development of six neural networks models for predicting slump and compressive strength of concrete with mineral additives such as blast furnace slag, fly ash and silica fume. The Back-Propagation Multi-Layer Perceptron (BPMLP) with Bayesian regularization was used in all these models. They are produced to implement the complex nonlinear relationship between the inputs and the output of the network. They are also established through the incorporation of a huge experimental database on concrete organized in the form Mix-Property. Thus, the data comprising the concrete mixtures are much correlated to each others. The PCA is proposed for the compression and the elimination of the correlation between these data. After applying the PCA, the uncorrelated data were used to train the six models. The predictive results of these models were compared with the actual experimental trials. The results showed that the elimination of the correlation between the input parameters using PCA improved the predictive generalisation performance models with smaller architectures and dimensionality reduction. This study showed also that using the developed models for numerical investigations on the parameters affecting the properties of concrete is promising.

• Smart Structures and Systems
• /
• 제14권2호
• /
• pp.159-189
• /
• 2014

One of the most important requirements in the evaluation of existing structural systems and ensuring a safe performance during their service life is damage assessment. Damage can be defined as a weakening of the structure that adversely affects its current or future performance which may cause undesirable displacements, stresses or vibrations to the structure. The mass and stiffness of a structure will change due to the damage, which in turn changes the measured dynamic response of the system. Damage detection can increase safety, reduce maintenance costs and increase serviceability of the structures. Artificial Neural Networks (ANNs) are simplified models of the human brain and evolved as one of the most useful mathematical concepts used in almost all branches of science and engineering. ANNs have been applied increasingly due to its powerful computational and excellent pattern recognition ability for detecting damage in structural engineering. This paper presents and reviews the technical literature for past two decades on structural damage detection using ANNs with modal parameters such as natural frequencies and mode shapes as inputs.

• 품질경영학회지
• /
• 제47권3호
• /
• pp.497-508
• /
• 2019

Purpose: The purpose of this study is to overcome limitations of conventional studies that to predict Baltic Dry Index (BDI). The study proposed applications of Artificial Neural Network (ANN) named Long Short-Term Memory (LSTM) to predict BDI. Methods: The BDI time-series prediction was carried out through eight variables related to the dry bulk market. The prediction was conducted in two steps. First, identifying the goodness of fitness for the BDI time-series of specific ANN models and determining the network structures to be used in the next step. While using ANN's generalization capability, the structures determined in the previous steps were used in the empirical prediction step, and the sliding-window method was applied to make a daily (one-day ahead) prediction. Results: At the empirical prediction step, it was possible to predict variable y(BDI time series) at point of time t by 8 variables (related to the dry bulk market) of x at point of time (t-1). LSTM, known to be good at learning over a long period of time, showed the best performance with higher predictive accuracy compared to Multi-Layer Perceptron (MLP) and Recurrent Neural Network (RNN). Conclusion: Applying this study to real business would require long-term predictions by applying more detailed forecasting techniques. I hope that the research can provide a point of reference in the dry bulk market, and furthermore in the decision-making and investment in the future of the shipping business as a whole.

• 제어로봇시스템학회논문지
• /
• 제14권8호
• /
• pp.818-823
• /
• 2008

In this paper, we propose a feature extraction method which extracts directional features of handwritten numerals by using the projection runlength. Our directional featrures are obtained from four directional images, each of which contains horizontal, vertical, right-diagonal and left-diagonal lines in entire numeral shape respectively. A conventional method which extracts directional features by using Kirsch masks generates edge-shaped double line directional images for four directions, whereas our method uses the projections and their runlengths for four directions to produces single line directional images for four directions. To obtain the directional projections for four directions from a numeral image, some preprocessing steps such as thinning and dilation are required, but the shapes of resultant directional lines are more similar to the numeral lines of input numerals. Four [$4{\times}4$] directional features of a numeral are obtained from four directional line images through a zoning method. By using a hybrid feature which is made by combining our feature with the conventional features of a mesh features, a kirsch directional feature and a concavity feature, higher recognition rates of the handwrittern numerals can be obtained. For recognition test with given features, we use a multi-layer perceptron neural network classifier which is trained with the back propagation algorithm. Through the experiments with the handwritten numeral database of Concordia University, we have achieved a recognition rate of 97.85%.

• 한국전산구조공학회:학술대회논문집
• /
• 한국전산구조공학회 2006년도 정기 학술대회 논문집
• /
• pp.1009-1014
• /
• 2006

Synthetic materials with excellent thermodynamic characteristics and the merit of anti-corrosion are frequently used in buildings and constructions for enforcement of bent in stead of steel plates. Among them, many practical studies have been conducted on bond strength because of increased bond strength of FRP plates. Previous investigators identified the bond strength of FRP plates through experiments with settlement of various variables to identify the bond strength. However, the experiments to identify the bond force are difficult to be conducted because they requires large expenses and long time for equipment arrangement, thus, are conducted with limitation. In this study, for bond experiment, optimum neural network model was developed with use of Back-propagation and Conjugate gradient technique of previous investigators. Learning was performed with use of the variables of previous investigators in developed neural network model so as to identify the bond strength of FRP plates. for verification of developed model, credibility and excellence was proven by comparing with the models of previous investigators.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제6권1호
• /
• pp.64-69
• /
• 2006

Due to the inherent nonlinear nature of Electro-rheological (ER) fluid dampers, one of the challenging aspects for utilizing these devices to achieve high system performance is the development of accurate models and control algorithms that can take advantage of their unique characteristics. In this paper, the nonlinear damping force model is made to identify the properties of the ER damper using higher order spectrum. The higher order spectral analysis is used to investigate the nonlinear frequency coupling phenomena with the damping force signal according to the sinusoidal excitation of the damper. Also, this paper presents an inverse model of the ER damper, i.e., the model can predict the required voltage so that the ER damper can produce the desired force for the requirement of vibration control of vehicle suspension systems. The inverse model is constructed by using a multi-layer perceptron neural network. A quarter-car suspension model is considered in this paper for analysis and simulation. Simulation results show that the proposed inverse model of ER damper can obtain control voltage of ER damper for required damping force.