• 정보학연구
• /
• 제4권3호
• /
• pp.155-163
• /
• 2001

디지털 신경회로망의 구현에 있어 시그모이드 함수의 구현은 매우 복잡하고 구현하기 어렵다. 따라서, 본 논문에서는 디지털 신경회로망 구현에 문제가 되는 시그모이드 함수처리를 위한 설계 방법을 제안하였다. 제안된 방법은 잉여수계를 이용하여 MAC(Multiplier and Accumulator) 연산 시, 캐리 전파 없이 고속의 연산을 수행할 수 있고 시그모이드 함수처리를 고속으로 수행할 수 있다. 모의실험결과, 각각의 신경 프로세스에 있어서 4.6nsec 이상의 속도를 보임으로써 고속디지털 신경회로망 구현에 적용될 수 있을 것으로 기대된다.

• International Journal of Control, Automation, and Systems
• /
• 제2권4호
• /
• pp.523-529
• /
• 2004

Sliding mode control with nonlinear interpolation in the boundary layer is proposed. A modified sigmoid function is used for nonlinear interpolation in the boundary layer and its parameter is tuned by a fuzzy controller. The fuzzy controller that takes both the sliding variable and a measure of chattering as its inputs tunes the parameter of the modified sigmoid function. Owing to the decreased thickness of the boundary layer and the tuned parameter, the proposed method has superior tracking performance than the conventional linear interpolation method.

• 한국산학기술학회논문지
• /
• 제15권3호
• /
• pp.1724-1733
• /
• 2014

본 논문에서는 비모노톤함수(non-monotone function)인 CosExp(cosine-modulated symmetric Exponential function) 함수와 모노톤함수(monotone function)인 시그모이드 함수를 캐스케이드 코릴레이션 알고리즘(Cascade Correlation algorithm)의 학습에 병행해서 사용하여 이중나선문제(two spirals problem)의 패턴인식에 어떠한 영향이 있는지 분석하고 이어서 알고리즘의 최적화를 시도한다. 첫 번째 실험에서는 알고리즘의 후보뉴런에 CosExp 함수를 그리고 출력뉴런에는 시그모이드 함수를 사용하여 나온 인식된 패턴을 분석한다. 두 번째 실험에서는 반대로 CosExp 함수를 출력뉴런에서 사용하고 시그모이드 함수를 후보뉴런에 사용하여 실험하고 결과를 분석한다. 세 번째 실험에서는 후보뉴런을 위한 8개의 풀을 구성하여 변형된 다양한 시그모이드 활성화 함수(sigmoidal activation function)를 사용하고 출력뉴런에는 CosExp함수를 사용하여 얻게 된 입력공간의 인식된 패턴을 분석한다. 네 번째 실험에서는 시그모이드 함수의 변위를 결정하는 세 개의 파라미터 값을 유전자 알고리즘을 이용하여 얻는다. 이 파라미터 값들이 적용된 시그모이드 함수들은 후보뉴런의 활성화를 위해서 사용되고 출력뉴런에는 CosExp 함수를 사용하여 실험한 최적화 된 결과를 분석한다. 이러한 알고리즘의 성능평가를 위하여 각 학습단계 마다 입력패턴공간에서 인식된 이중나선의 형태를 그래픽으로 보여준다. 최적화 과정에서 은닉뉴런(hidden neuron)의 숫자가 28에서 15로 그리고 최종적으로 12개로 줄어서 학습 알고리즘이 최적화되었음을 확인하였다.

• 디지털산업정보학회논문지
• /
• 제19권4호
• /
• pp.137-146
• /
• 2023

The dropout of university freshmen is a very important issue in the financial problems of universities. Moreover, the dropout rate is one of the important indicators among the external evaluation items of universities. Therefore, universities need to predict dropout students in advance and apply various dropout prevention programs targeting them. This paper proposes a method to predict such dropout students in advance. This paper is about a method for predicting dropout students. It proposes a method to select dropouts by applying logistic regression using a shift sigmoid classification function using only quantitative data from the first semester of the first year, which most universities have. It is based on logistic regression and can select the number of prediction subjects and prediction accuracy by using the shift sigmoid function as an classification function. As a result of the experiment, when the proposed algorithm was applied, the number of predicted dropout subjects varied from 100% to 20% compared to the actual number of dropout subjects, and it was found to have a prediction accuracy of 75% to 98%.

• Journal of Applied and Pure Mathematics
• /
• 제4권3_4호
• /
• pp.185-209
• /
• 2022

Here we exhibit multivariate quantitative approximations of Banach space valued continuous multivariate functions on a box or ℝ^N, N ∈ ℕ, by the multivariate normalized, quasi-interpolation, Kantorovich type and quadrature type neural network operators. We treat also the case of approximation by iterated operators of the last four types. These approximations are achieved by establishing multidimensional Jackson type inequalities involving the multivariate modulus of continuity of the engaged function or its high order Fréchet derivatives. Our multivariate operators are defined by using a multidimensional density function induced by the generalized symmetrical sigmoid function. The approximations are point-wise and uniform. The related feed-forward neural network is with one hidden layer.

• Structural Engineering and Mechanics
• /
• 제74권1호
• /
• pp.83-90
• /
• 2020

Several classical and higher order plate theories were used to study the buckling of functionally graded material (FGM) plates. In the great majority of research, a power function is used to represent metal and ceramic material transverse distribution (P-FGM). Therefore, the effect of having other transverse variation of material properties on the buckling behavior of thick rectangular FGM plates was not properly addressed. In the present work, this effect is investigated using the Third order Shear Deformable Theory (TSDT) for the case of simply supported FGM plate. Both a sigmoid function and an exponential functions are used to represent the transverse gradual property variation. The plate governing equations are combined with a Navier type expanded solution of the unknown displacements to derive the buckling equation in terms of the pre-buckling in-plane loads. Finally, the critical in-plane load is calculated for the different buckling modes. The model is verified by a comparison of the calculated buckling loads with available published results of Al-SiC P-FGM plates. The conducted parametric study shows that manufacturing FGM plates with sigmoid variation of properties in the thickness direction increases the buckling load considerably. This improvement is found to be more significant for the case of thick plates than that of thin plates. Results also show that this stiffening-like effect of the sigmoid function profile is more evident for cases where the in-plane loads are applied along the shorter edge of the plate.

• Journal of Information Processing Systems
• /
• 제13권4호
• /
• pp.703-715
• /
• 2017

Mammogram images are sensitive in nature and even a minor change in the environment affects the quality of the images. Due to the lack of expert radiologists, it is difficult to interpret the mammogram images. In this paper an algorithm is proposed for a computer-aided diagnosis system, which is based on the wavelet based adaptive sigmoid function. The cascade feed-forward back propagation technique has been used for training and testing purposes. Due to the poor contrast in digital mammogram images it is difficult to process the images directly. Thus, the images were first processed using the wavelet based adaptive sigmoid function and then the suspicious regions were selected to extract the features. A combination of texture features and gray-level co-occurrence matrix features were extracted and used for training and testing purposes. The system was trained with 150 images, while a total 100 mammogram images were used for testing. A classification accuracy of more than 95% was obtained with our proposed method.

• 전기전자학회논문지
• /
• 제22권1호
• /
• pp.197-200
• /
• 2018

본 논문에서는 하드웨어레벨로 구현이 어렵고 속도가 느린 sigmoid 함수를 PLAN을 이용하여 근사치로 출력하였다. 이를 MLP 구조의 활성화 함수로 사용하여 자원소모를 줄이고 속도를 개선하고자 하였다. 본 논문에서 제안하는 방법은 $5{\times}5$크기의 숫자 인식에 약 95%의 정확도를 유지하면서 GPGPU보다 약 1.83배의 빠른 속도를 보였다. 또한 MLPA가속기와 비슷한 자원을 사용함에도 더 많은 뉴런을 사용하여 높은 정확도에 빠른 속도로 수렴하는 것을 확인하였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2000년도 제15차 학술회의논문집
• /
• pp.20-20
• /
• 2000

Back-propagation neural network (BPNN) is the most prevalently used paradigm in modeling semiconductor manufacturing processes, which as a neuron activation function typically employs a bipolar or unipolar sigmoid function in either hidden and output layers. In this study, applicability of another linear function as a neuron activation function is investigated. The linear function was operated in combination with other sigmoid functions. Comparison revealed that a particular combination, the bipolar sigmoid function in hidden layer and the linear function in output layer, is found to be the best combination that yields the highest prediction accuracy. For BPNN with this combination, predictive performance once again optimized by incrementally adjusting the gradients respective to each function. A total of 121 combinations of gradients were examined and out of them one optimal set was determined. Predictive performance of the corresponding model were compared to non-optimized, revealing that optimized models are more accurate over non-optimized counterparts by an improvement of more than 30%. This demonstrates that the proposed gradient-optimized teaming for BPNN with a linear function in output layer is an effective means to construct plasma models. The plasma modeled is a hemispherical inductively coupled plasma, which was characterized by a 24 full factorial design. To validate models, another eight experiments were conducted. process variables that were varied in the design include source polver, pressure, position of chuck holder and chroline flow rate. Plasma attributes measured using Langmuir probe are electron density, electron temperature, and plasma potential.

• Journal of applied mathematics & informatics
• /
• 제23권1_2호
• /
• pp.563-570
• /
• 2007

There has been a considerable research interest in medical communities for neuronal fiber tracking with magnetic resonance diffusion tensor imaging(DTI). DTI data have abundant structural boundaries that need to be preserved during interpolation to facilitate fiber tracking. Sigmoid function has been used in recent papers but the sigmoid function still is not good enough to be served as an positive interpolation in mathematical point of view. In this paper, we construct and provide two families positive cardinal interpolation functions.