• Journal of KIISE:Software and Applications
• /
• v.31 no.5
• /
• pp.577-585
• /
• 2004

Storing and estimating the high order probability distribution of classifiers and class labels is exponentially complex and unmanageable without an assumption or an approximation, so we rely on an approximation scheme using the dependency. In this paper, as an extended study of the second-order dependency-based approximation, the probability distribution is optimally approximated by the third-order dependency. The proposed third-order dependency-based approximation is applied to the combination of multiple classifiers recognizing handwritten numerals from Concordia University and the University of California, Irvine and its usefulness is demonstrated through the experiments.

• The Korean Journal of Applied Statistics
• /
• v.22 no.1
• /
• pp.171-179
• /
• 2009

In this paper we studied the effective approximations to the distribution of the sum of products of normal variables. Based on the saddlepoint approximations to the quadratic forms, the suggested approximations are very accurate and easy to use. Applications to the FSK (Frequency Shift Keying) communication are also considered.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.4
• /
• pp.671-678
• /
• 2022

Approximate computing is an computational technique that is acceptable degree of inaccurate results of accurate results. Approximate multiplication is one of the approximate computing methods for high-performance and low-power computing. In this paper, we propose a high-density, low-power, and high-speed approximate multiplier using approximate 4-2 compressor and improved full adder. The approximate multiplier with approximate 4-2 compressor consists of three regions of the exact, approximate and constant correction regions, and we compared them by adjusting the size of region by applying an efficient partial product reduction. The proposed approximate multiplier was designed with Verilog HDL and was analyzed for area, power and delay time using Synopsys Design Compiler (DC) on a 25nm CMOS process. As a result of the experiment, the proposed multiplier reduced area by 10.47%, power by 26.11%, and delay time by 13% compared to the conventional approximate multiplier.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.1
• /
• pp.173-180
• /
• 2022

Approximate Computing is a promising method for designing hardware-efficient computing systems. Approximate multiplication is one of key operations used in approximate computing methods for high performance and low power computing. An approximate 4-2 compressor can implement hardware-efficient circuits for approximate multiplication. In this paper, we propose an approximate multiplier with low area and low power characteristics. The proposed approximate multiplier architecture is segmented into three portions; an exact region, an approximate region, and a constant correction region. Partial product reduction in the approximation region are simplified using a new 4:2 approximate compressor, and the error due to approximation is compensated using a simple error correction scheme. Constant correction region uses a constant calculated with probabilistic analysis for reducing error. Experimental results of 8×8 multiplier show that the proposed design requires less area, and consumes less power than conventional 4-2 compressor-based approximate multiplier.

• Journal of KIISE:Software and Applications
• /
• v.27 no.8
• /
• pp.855-863
• /
• 2000

Although Behavior-Knowledge Space (BKS) method, one of well known decision combination methods, does not need any assumptions in combining the multiple experts, it should theoretically build exponential storage spaces for storing and managing jointly observed K decisions from K experts. That is, combining K experts needs a (K+1)st-order probability distribution. However, it is well known that the distribution becomes unmanageable in storing and estimating, even for a small K. In order to overcome such weakness, it has been studied to decompose a probability distribution into a number of component distributions and to approximate the distribution with a product of the component distributions. One of such previous works is to apply a conditional independence assumption to the distribution. Another work is to approximate the distribution with a product of only first-order tree dependencies or second-order distributions as shown in [1]. In this paper, higher order dependency than the first-order is considered in approximating the distribution and a dependency-based framework is proposed to optimally approximate the (K+1)st-order probability distribution with a product set of dth-order dependencies where ($1{\le}d{\le}K$), and to combine multiple experts based on the product set using the Bayesian formalism. This framework was experimented and evaluated with a standardized CENPARMI data base.

• Journal of KIISE:Software and Applications
• /
• v.28 no.9
• /
• pp.681-687
• /
• 2001

In order to raise a class discrimination power by combining multiple classifiers under the Bayesian decision theory, the upper bound of a Bayes error rate bounded by the conditional entropy of a class variable and decision variables obtained from training data samples should be minimized. Wang and Wong proposed a tree dependence first-order approximation scheme of a high order probability distribution composed of the class and multiple feature pattern variables for minimizing the upper bound of the Bayes error rate. This paper presents an extended high order product approximation scheme dealing with higher order dependency more than the first-order tree dependence, based on the minimization of the upper bound of the Bayes error rate. Multiple recognizers for unconstrained handwritten numerals from CENPARMI were combined by the proposed approximation scheme using the Bayesian formalism, and the high recognition rates were obtained by them.

• The Journal of the Acoustical Society of Korea
• /
• v.12 no.1
• /
• pp.20-26
• /
• 1993

본 논문은 Reed-Solomon부호의 복호가능어 가중치 분포에 대한 명시적 식과 근사식을 구하여 이를 복호기 오류확률 PE(u)에 적용하고, 복호기 오류확률의 상한식을 구하고 분석하였다. t+1개 이상의 오류가 발생했을 때 복호기 오류확률의 추정치 Q와 Q'를 개선하여 식 Q를 제안하고, 컴퓨터 시뮬레이션을 수행한 결과 가중치 u가 커질 때 복호기 오류확률은 추정치 Q와 Q'에는 접근하였으나, 본 논문에서 제안한 Q와는 일치됨을 확인하였다. 그리고, 가중치 u가 부호의 길이 n에 접근할 때, 복호가능어의 명시적 식 Du와 근사식 Du'가 서로 일치하고, 복호기 오류확률 Pe(u)와 근사오류확률 Pe(u')가 일치함을 보였다. 또하 t+1개 이상의 오류가 발생했을 때 복호기 오류확률은 1/t!보다 작으며, 가중치분포 Au에 Vn(t)를 곱한 결과는 근사복호가능어 Du'와 일치함도 확인하였다.

• The Journal of the Acoustical Society of Korea
• /
• v.36 no.1
• /
• pp.1-11
• /
• 2017

In this study, novel approximate form of the square-root operator of three dimensional acoustic Parabolic Equation (3D PE) is proposed using a rational approximant for two variables. This form has two advantages in comparison with existing approximation studies of the square-root operator. One is the wide-angle capability. The proposed form has wider angle accuracy to the inclination angle of ${\pm}62^{\circ}$ from the range axis of 3D PE at the bearing angle of $45^{\circ}$, which is approximately three times the angle limit of the existing 3D PE algorithm. Another is that the denominator of our approximate form can be expressed into the product of one-dimensional operators for depth and cross-range. Such a splitting form is very preferable in the numerical analysis in that the 3D PE can be easily transformed into the tridiagonal matrix equation. To confirm the capability of the proposed approximate form, comparative study of other approximation methods is conducted based on the phase error analysis, and the proposed method shows best performance.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.280-282
• /
• 2020

최근 사물인터넷(IoT), 자율주행과 같이 기계 간의 통신이 요구되는 서비스가 늘어감에 따라, 기계 임무 수행에 최적화된 데이터의 생성 및 압축에 대한 필요성이 증가하고 있다. 또한, 사물인터넷과 인공지능(AI)이 접목된 기술이 주목을 받으면서 딥러닝 모델에서 추출되는 특징(feature)을 디바이스에서 클라우드로 전송하는 방안에 관한 연구가 진행되고 있으며, 국제 표준화 기구인 MPEG에서는 '기계를 위한 부호화(Video Coding for Machine: VCM)'에 대한 표준 기술 개발을 진행 중이다. 딥러닝으로 특징을 추출하는 가장 대표적인 방법으로는 합성곱 신경망(Convolutional Neural Network: CNN)이 있으며, 오토인코더는 입력층과 출력층의 구조를 동일하게 하여 출력을 가능한 한 입력에 근사시키고 은닉층을 입력층보다 작게 구성하여 차원을 축소함으로써 데이터를 압축하는 딥러닝 기반 이미지 압축 방식이다. 이에 본 논문에서는 이러한 오토인코더의 성질을 이용하여 CNN 기반의 이미지 분류 네트워크의 합성곱 신경망으로부터 추출된 feature에 오토인코더를 적용하여 압축하는 방안을 제안한다.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.36 no.10
• /
• pp.1109-1114
• /
• 2012

The performance of a system can be affected by various variables such as manufacturing tolerances, uncertainties of material properties, and environmental factors acting on the system. Robust design optimization has attracted much attention in the design of products because it can find the best design solution that minimizes the variance of the response while considering the distribution of the variables. However, the computational cost and accuracy of optimization have thus far been a challenging problem. In this study, robust design optimization using the multiplicative decomposition method is proposed in order to solve these problems. Because the proposed method calculates the mean and variance of the system directly from the kriging metamodel using the multiplicative decomposition method, it can be used to search for a robust optimum design accurately and efficiently. Several mathematical and engineering examples are used to demonstrate the feasibility of the proposed method.