• 전자공학회논문지C
• /
• 제35C권3호
• /
• pp.79-89
• /
• 1998

The Error Back-Propagation(EBP) algorithm is widely applied to train a multi-layer perceptron, which is a neural network model frequently used to solve complex problems such as pattern recognition, adaptive control, and global optimization. However, the EBP is basically a gradient descent method, which may get stuck in a local minimum, leading to failure in finding the globally optimal solution. Moreover, a multi-layer perceptron suffers from locking a systematic determination of the network structure appropriate for a given problem. It is usually the case to determine the number of hidden nodes by trial and error. In this paper, we propose a new algorithm to efficiently train a multi-layer perceptron. OUr algorithm uses stochastic perturbation in the weight space to effectively escape from local minima in multi-layer perceptron learning. Stochastic perturbation probabilistically re-initializes weights associated with hidden nodes to escape a local minimum if the probabilistically re-initializes weights associated with hidden nodes to escape a local minimum if the EGP learning gets stuck to it. Addition of new hidden nodes also can be viewed asa special case of stochastic perturbation. Using stochastic perturbation we can solve the local minima problem and the network structure design in a unified way. The results of our experiments with several benchmark test problems including theparity problem, the two-spirals problem, andthe credit-screening data show that our algorithm is very efficient.

• 대한기계학회:학술대회논문집
• /
• 대한기계학회 2004년도 추계학술대회
• /
• pp.858-863
• /
• 2004

In the optimized design of an actual structure, the design variable should be selected among any certain values or corresponds to a discrete design variable that needs to handle the size of a pre-formatted part. Various algorithms have been developed for discrete design. As recently reported, the sequential algorithm with orthogonal arrays(SOA), which is a local minimum search algorithm in discrete space, has excellent local minimum search ability. It reduces the number of function evaluation using orthogonal arrays. However it only finds a local minimum and the final solution depends on the initial value. In this research, the genetic algorithm, which defines an initial population with the potential solution in a global space, is adopted in SOA. The new algorithm, sequential algorithm with orthogonal arrays and genetic algorithm(SOAGA), can find a global solution with the properties of genetic algorithm and the solution is found rapidly with the characteristics of SOA.

• Journal of Welding and Joining
• /
• 제21권7호
• /
• pp.42-48
• /
• 2003

Welding distortion is a current issue in many industrial parts, especially for heavy industry such as shipbuilding, plant industry. The welding process has many processing parameters influencing welding angular distortion such as heat input power, welding speed, gas flow rate, plate thickness and the welded material properties, etc. In this work, the conventional local minimization concept was applied to find a set of optimum welding process parameters, consisted of welding speed, plate thickness and heat input, for a minimum angular distortion. An analytic solution for welding angular distortion, which is based on laminated plate theory, was also applied to investigate and optimize the welding process parameters. The optimized process parameters and the angular distortion for various parametric conditions could be easily found by using the local minimum concept.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1999년도 하계학술대회 논문집 G
• /
• pp.3086-3088
• /
• 1999

An ultrasonic sensor is one of most popular sensor used to navigate mobile robots within environments containing obstacles. But many navigation algorithm have studied because of the drawback of ultrasonic sensor such that poor directionality, frequent misreadings, specular reflections. Also, the most crucial drawback of this algorithm, that is VFF, VFM, EDM, PFM, WFM, GFM etc. has been that the mobile robot may become trapped in a local minimum. In this paper, we present a theoretical study of a navigation algorithm which integrals a heuristic-search local minimum (or trap) recovery method with a vector-field based method to maneuver cylindric mobile robots in unknown of unstructured environments. Also, an autonomous mobile robot uses dead-reckoning to estimate the current position and orientation of a mobile robot.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2009년도 추계학술발표대회
• /
• pp.295-296
• /
• 2009

본 논문은 기존 PSO의 문제점으로 지적되는 Local minimum에의 고착을 해결하여 평균적인 성능을 향상시킬 수 있는 방법을 제안한다. 이를 위해 기존에는 하나의 Best Value만을 향해서 이동하는 것에서 벗어나, 각 Particle들을 Cluster로 나누고 각각의 Best Value를 중심으로 이동하며, 각 Cluster Leader들은 다른 Cluster Leader와 통신하며 순위에 따라 이동한다. 제안 방법으로 수렴속도와 Local minimum 회피에 대해 실험을 통해 비교 분석한다. 실험 결과로부터 제안 방법이 Local minimum 회피에서 성능 향상이 있음을 고찰하였다.

• Journal of Astronomy and Space Sciences
• /
• 제23권2호
• /
• pp.117-126
• /
• 2006

지상에서 우주선 측정기(NM, Neutron Monitor)로 관측되는 우주선 세기의 일변화는 일평균 값의 $1{\sim}2%$ 정도의 진폭을 갖는 사인곡선의 형태를 보인다. 본 연구는 NM 관측소에서 우주선 일변화의 최대 및 최소 세기가 관측되는 지방시 변화의 통계적 경향 분석을 수행하였다. 우주선 일변화 최대 및 최소 세기 시각의 분포에 영향을 주는 우주선 중성자 관측소의 위치(cut-off rigidity)와 태양활동도의 영향을 분석해 보기 위해서 저위도 지방의 Haleakala(위도: 20.72N, cut-off rigidity: 12.91Gev)와 고위도 지방의 Oulu(위도: 65.05N, cut-off rigidity: 0.81Gev) 우주선 중성자 관측소의 1996년(태양활동 극소기) 및 2000년(태양활동 극대기) 우주선 관측 자료를 분석하였다. 태양활동 극대기의 우주선 일변화 최대 및 최소 세기 시각은 태양활동 극소기에 비해 약 $2{\sim}3$시간 정도 늦은 시각에 나타난다. 우주선 중성자 관측소의 위치 즉 위도별 분포(rigidity의 크기)에 따른 영향을 살펴보면, cut-off. rigidity가 Haleakala보다 작은 Oulu에서 우주선 일변화 최대 및 최소 세기 시각이 약 $2{\sim}3$시간 정도 늦게 나타나며, 이러한 현상은 극대기에 더욱 뚜렷하게 나타났다. 태양활동에 따른 행성간 자기장 세기의 변화와 위도에 따른 cut-off rigidity의 크기에 따라 우주선 세기 일변화 위상이 결정된다고 볼 수 있다.

• 대한기계학회논문집
• /
• 제19권4호
• /
• pp.887-895
• /
• 1995

Considering that the laser range finder has the excellent resolution with respect to angular and distance measurements, a sophisticated local path planning algorithm is achieved by subgoal and sub-subgoal searching methods. The subgoal searching finds the passable ways between obstacles and selects the optimal pathway in order to reduce the moving distanced from start point to given to given goal. On the other hand, the sub-subgoal searching corrects the path given in subgoal searching in the case of which the mobile robot will collide with obstacles. Also, the effectiveness of the established local path planning and local minimum avoiding algorithm are estimated by computer simulation and experimentation in complex environment.

• 한국복합재료학회:학술대회논문집
• /
• 한국복합재료학회 2000년도 춘계학술발표대회 논문집
• /
• pp.9-14
• /
• 2000

This paper presents the analytical results of local buckling of orthotropic I-shape compression members. Employing the equilibrium approach, the characteristic equation for local buckling of I-shape compression member is derived. Using the derived equation, the minimum buckling coefficients with respect to the ratio of width to thickness for the I-shape column are suggested as a graphical form. In addition, the dominant plate component initiating the local buckling of I-shape column is also identified by using the approximate solution and the results are plotted with dotted line on the minimum bucking coefficient curve.

• 천문학회보
• /
• 제36권2호
• /
• pp.99-99
• /
• 2011

A statistical study of coronal hole merging and splitting has been performed through Solar Cycle 23. The NOAA/SESC solar synoptic maps are examined to identify inarguably clear events of coronal hole merging and splitting. The numbers of merging events and splitting events are more or less comparable regardless of the phase in the solar cycle. The number of both events, however, definitely shows the phase dependence in the solar cycle. It apparently has a minimum at the solar minimum whereas its maximum is located in the declining phase of the sunspot activity, about a year after the second peak in Solar Cycle 23. There are more events of merging and splitting in the descending phase than in the ascending phase. Interestingly, no event is found at the local minimum between the two peaks of the sunspot activity. This trend can be compared with the variation of the average magnetic field strength and the radial field component in the solar wind through the solar cycle. In Ulysses observations, both of these quantities have a minimum at the solar minimum while their maximum is located in the descending phase, a while after the second peak of the sunspot activity. At the local minimum between the two peaks in the solar cycle, the field strength and the radial component both have a shallow local minimum or an inflection point. At the moment, the physical reason for these resembling tendencies is difficult to understand with existing theories. Seeing that merging and splitting of coronal holes are possible by passage of opposite polarity magnetic structures, we may suggest that the energizing activities in the solar surface such as motions of flux tubes are not exactly in phase with sunspot generation, but are more active some time after the sunspot maximum.

• 대한기계학회논문집
• /
• 제16권4호
• /
• pp.630-648
• /
• 1992

본 연구에서는 여러가지 하중조건하에서 단순지지된 보강원통셸의 최소중량화 설계문제를 CONMIN을 사용하여 해석하고, 일반적인 대칭적층[0/.+-..theta./90]$_{s}$ 의 복합 적층원통셸, 복합적층honeycomb sandwich원통셸, 그리고 보강된 복합적층원통셸의 최 소중량화 설계문제에도 확장 적용한다. 설계변수(design variable)로는 등방성재료 인 경우와 복합적층인 경우 최대 9개, 부등제한조건으로는 전체좌굴(general buckling ), 준전체좌굴(panel buckling), 판 및 보강재의 국부좌굴(local cripping), 로링모드 (rolling mode), 그리고 응력과 변형률제한 등의 성질제한조건(behavior constraints) 과 설계변수의 상, 하한을 나타내는 기하학적 제한(side constraints)등 최대 32개를 설정한다. 본 최소중량화 설계예에서는 보강재의 최적단면형상을 검토하기 위하여 직사각형(R)형, I형, 그리고 T형 단면 등의 보강재들을 사용한다.