• Journal of the Korea Society of Computer and Information
• /
• v.27 no.9
• /
• pp.69-76
• /
• 2022

The traveling salesman problem(TSP) is one of the most famous combinatorial optimization problem. So far, many metaheuristic search algorithms have been proposed to solve the problem, and one of them is local search. One of the very important factors in local search is neighbor generation method, and random-based neighbor generation methods such as inversion have been mainly used. This paper proposes 4 new greedy-based neighbor generation methods. Three of them are based on greedy insertion heuristic which insert selected cities one by one into the current best position. The other one is based on greedy rotation. The proposed methods are applied to first-choice hill-climbing search and simulated annealing which are representative local search algorithms. Through the experiment, we confirmed that the proposed greedy-based methods outperform the existing random-based methods. In addition, we confirmed that some greedy-based methods are superior to the existing local search methods.

• Journal of KIISE
• /
• v.45 no.2
• /
• pp.157-164
• /
• 2018

Due to the revolutionary advances in deep learning, performance of pattern recognition has increased significantly in many applications like speech recognition and image recognition, and some systems outperform human-level intelligence in specific domains. Unlike pattern recognition, in this paper, we focus on generating Korean sentences based on a few Korean sentences. We apply variational recurrent auto-encoder (VRAE) and modify the model considering some characteristics of Korean sentences. To reduce the number of words in the model, we apply a word spacing model. Also, there are many Korean sentences which have the same meaning but different word order, even without subjects or objects; therefore we change the unidirectional encoder of VRAE into a bidirectional encoder. In addition, we apply an interpolation method on the encoded vectors from the given sentences, so that we can generate new sentences which are similar to the given sentences. In experiments, we confirm that our proposed method generates better sentences which are semantically more similar to the given sentences.

• KIPS Transactions on Software and Data Engineering
• /
• v.6 no.6
• /
• pp.293-302
• /
• 2017

This paper proposes a method that generates test cases from Simulink/Stateflow(SL/SF) using a SMT (Satisfiability Modulo Theory) solver, Yices and information of SL/SF model. The most difficult problem to generate test cases from SL/SF model is to solve reachability problem. In the propose method, Yices and the tables built with the model information are utilized to solve the reachability problem. The method utilizes the SMT model, that is the SL/SF model transformed in Yices. The tables built from SL/SF are used for backward processing of the proposed method and increases test generation efficiency. A commercial refrigerator model and two car ECU (Electrical Control Unit) models are used to evaluate the performance of the proposed algorithm..

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 1999.03a
• /
• pp.271-280
• /
• 1999

인공신경망을 시계열예측에 적용하는 경우에 고려되어야 할 문제중, 특히 모형에 적합한 입력변수의 생성이 중요시되고 있는데, 이러한 분야는 인공신경망의 모형생성과정에서 입력변수에 대한 전처리기법으로써 다양하게 제시되어 왔다. 가장 최근의 입력변수 전처리기법으로써 제시되고 있는 신호처리기법은 전통적 주기분할처리방법인 푸리에변환기법(Fourier transforms)을 비롯하여 이를 확장시킨 개념인 웨이블릿변환기법(wavelet transforms) 등으로 대별될 수 있다. 이는 기본적으로 시계열이 다수의 주기(cycle)들로 구성된 상이한 시계열들의 집합이라는 가정에서 출발하고 있다. 전통적으로 이러한 시계열은 전기 또는 전자공학에서 주파수영역분할, 즉 고주파 및 저주파수를 분할하기 위한 기법에 적용되어 왔다. 그러나, 최근에는 이러한 연구가 다양한 분야에 활발하게 응용되기 시작하였으며, 그 중의 대표적인 예가 바로 경영분야의 재무시계열에 대한 분석이다. 전통적으로 재무시계열은 장, 단기의사결정을 가진 시장참여자들간의 거래특성이 시계열에 각기 달리 가격으로 반영되기 때문에 이러한 상이한 집단들의 고요한 거래움직임으로 말미암아 예를 들어, 주식시장이 프랙탈구조를 가지고 있다고 보기도 한다. 이처럼 재무시계열은 다양한 사회현상의 집합체라고 볼 수 있으며, 그만큼 예측모형을 구축하는데 어려움이 따른다. 본 연구는 이러한 시계열의 주기적 특성에 기반을 둔 신호처리분석으로서 기존의 시계열로부터 노이즈를 줄여 주면서 보다 의미있는 정보로 변환시켜줄 수 있는 웨이블릿분석 방법론을 새로운 필터링기법으로 사용하여 현재 많은 연구가 진행되고 있는 인공신경망의 모형결합을 통해 기존연구과는 다른 새로운 통합예측방법론을 제시하고자 한다. 본 연구에서는 제시하는 통합방법론은 크게 2단계 과정을 거쳐 예측모형으로 완성이 된다. 즉, 1차 모형단계에서 원시 재무시계열은 먼저 웨이브릿분석을 통해서 노이즈가 필터링 되는 동시에, 과거 재무시계열의 프랙탈 구조, 즉 비선형적인 움직임을 보다 잘 반영시켜 주는 다차원 주기요소를 가지는 시계열로 분해, 생성되며, 이렇게 주기에 따라 장단기로 분할된 시계열들은 2차 모형단계에서 신경망의 새로운 입력변수로서 사용되어 최종적인 인공 신경망모델을 구축하는 데 반영된다. 기존의 주기분할방법론은 모형개발자입장에서 여러 가지 통계기준치중에서 최적의 기준치를 합리적으로 선택해야 하는 문제가 추가적으로 발생하며, 본 연구에서는 이상의 제반 문제들을 개선시키기 위해 통합방법론으로서 기존의 인공신경망모형을 구조적으로 확장시켰다. 이 모형에서 기존의 입력층 이전단계에 새로운 층이 정의된다. 이렇게 해서 생성된 새로운 통합모형은 기존모형에서 생성되는 기본적인 학습파라미터와 더불어, 본 연구에서 새롭게 제시된 주기분할층의 파라미터들이 모형의 학습성과를 높이기 위해 함께 고려된다. 한편, 이러한 학습과정에서 추가적으로 고려해야 할 파라미터 갯수가 증가함에 따라서, 본 모델의 학습성과가 local minimum에 빠지는 문제점이 발생될 수 있다. 즉, 웨이블릿분석과 인공신경망모형을 모두 전역적으로 최적화시켜야 하는 문제가 발생한다. 본 연구에서는 이 문제를 해결하기 위해서, 최근 local minimum의 가능성을 최소화하여 전역적인 학습성과를 높여 주는 인공지능기법으로서 유전자알고리즘기법을 본 연구이 통합모델에 반영하였다. 이에 대한 실증사례 분석결과는 일일 환율예측문제를 적용하였을 경우, 기존의 방법론보다 더 나운 예측성과를 타나내었다.

• Journal of the Korean Chemical Society
• /
• v.56 no.5
• /
• pp.648-657
• /
• 2012

This study is examined for the effects of question generating strategy and feedback(teacher feedback and student peer reviews) on science achievement, self-efficiency and perception of the class in the middle school. Three classes of middle school 1st grade in a city were sampled for the study. The students in comparative group took traditional lessons and solved questions presented on worksheets distributed by a teacher. On the other hand, the students in the experimental groups 1 and 2 generated and solved questions by themselves after traditional lessons and then participated in the activity with peer's feedback and teacher's feedback, respectively. A self-efficiency test had been conducted before the treatments and the various tests such as achievement, self-efficiency and the perception of the class were carried out after treatments. The statistical results were analysed by ANCOVA, MANCOVA and t-test. The results of this study were as follows: First, there was significant difference between the control and experimental groups in the science achievement test (p<.01). But there was no significant difference between the experimental groups (p>.05). Second, both of the experimental groups showed enhanced self-efficiency compared with comparative group (p<.01). In sub-areas of self-efficiency, there were also meaningful results in the area of self-efficiency for control experimental group 1(p<.01), and in the area of difficulty preferences for experimental group 2 (p<.01). Third, the experimental group 1, given teacher's feedback, showed more positive perception of the class than the experimental group 2 that was given peer's feedback (p<.01).

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2005.10a
• /
• pp.57-78
• /
• 2005

The vehicle routing problem (VRP) is to determine a set of feasible vehicle routes, one for each vehicle, such that each customer is visited exactly once and the total distance travelled by the vehicles is minimized. A feasible route is defined as a simple circuit including the depot such that the total demand of the customers in the route does not exceed the vehicle capacity. While there have been significant advances recently in exact solution methodology, the VRP is not a well solved problem. We find most approaches still relying on the branch and bound method. These approaches employ various methodologies to compute a lower bound on the optimal value. We introduce a new modelling approach, termed route-splitting, for the VRP that allows us to address problems whose size is beyond the current computational range of set-partitioning models. The route-splitting model splits each vehicle route into segments, and results in more tractable subproblems. Lifting much of the burden of solving combinatorially hard subproblems, the route-splitting approach puts more weight on the LP master problem, Recent breakthroughs in solving LP problems (Nemhauser, 1994) bode well for our approach. Lower bounds are computed on five symmetric VRPs with up to 199 customers, and eight asymmetric VRPs with up to 70 customers. while it is said that the exact methods developed for asymmetric instances have in general a poor performance when applied to symmetric ones (Toth and Vigo, 2002), the route splitting approach shows a competent performance of 93.5% on average in the symmetric VRPs. For the asymmetric ones, the approach comes up with lower bounds of 97.6% on average. The route-splitting model can deal with asymmetric cost matrices and non-identical vehicles. Given the ability of the route-splitting model to address a wider range of applications and its good performance on asymmetric instances, we find the model promising and valuable for further research.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2003.05a
• /
• pp.129-136
• /
• 2003

설비입지문제는 고객에게 좋은 서비스를 제공하면서 전체 비용을 최소화하는 의사결정을 요구한다. 본 논문은 제한된 총 투자 비용 하에서 최대의 서비스 수준을 달성하기 위하여 고정비를 가지는 설비의 입지를 결정하는 문제에 관한 것이다. 이 문제에 대해 수리 모형을 제시하고, 라그랑지안 기법을 이용한 발견적 기법을 통하여 해를 구하였다. 문제의 상한(Upper Bound)은 서브그래디언트(Subgradient) 최적화 기법을 사용하여 구하였고, 하한(Lower Bound)을 구하기 위하여 커팅(Cutting) 알고리즘이라는 새로운 기법을 개발하여 적용하였다. 임의로 생성된 데이터를 이용하여 비용과 커버 가능거리라는 두 가지 관점에서 실험을 하고 제안된 알고리즘의 성능을 비교 분석하였다.

• Journal of Information Management
• /
• v.41 no.2
• /
• pp.31-46
• /
• 2010

Now named-entity recognition(NER) as a part of information extraction has been used in the fields of information retrieval as well as question-answering systems. Unlike words, named-entities(NEs) are generated and changed steadily in documents on the Web, newspapers, and so on. The NE generation causes an unknown word problem and makes many application systems with NER difficult. In order to alleviate this problem, this paper proposes a new feature generation method for machine learning-based NER. In general features in machine learning-based NER are related with words, but entities in named-entity dictionaries are related to phrases. So the entities are not able to be directly used as features of the NER systems. This paper proposes an encoding scheme as a feature generation method which converts phrase entities into features of word units. Futhermore, due to this scheme, entities with semantic information in WordNet can be converted into features of the NER systems. Through our experiments we have shown that the performance is increased by about 6% of F1 score and the errors is reduced by about 38%.

• Lingua Humanitatis
• /
• v.6
• /
• pp.77-108
• /
• 2004

This research focuses on the humboldt's thought for language. In Korea, only two themes, 'worldview' and 'nation' , problem of 'energeia' and 'ergon', was mostly researched. But we must research all together his thoughts, in order to understand his complete thoughts. His thoughts consist six themes : 1 . relation between thought and language. 2. problem of language-generation. 3. problem of translation. 4. substance of language. 5. worldview and nation. 6. energeia and ergon. From this five themes, we can understand, what his origin thought is. Therefore, we explained sequentially five themes in our paper. The theme 1, 5 and 6 are often interpretated between the scholars in Korea. But the theme 2 and 3 are not yet known. Therefore, we tried to explain it in this paper, what the problem of language-generation and the problem of translation are.tion and the problem of translation are.

• Journal of Intelligence and Information Systems
• /
• v.14 no.4
• /
• pp.1-18
• /
• 2008

By using previous similar case plans, the case-based planning (CBP) systems can generate efficiently plans for new problems. However, most existing CBP systems show limited functionalities for case retrieval and case generalization. Moreover, they do not allow their users to participate in the process of plan generation. To support efficient memory use and case retrieval, the proposed case-based planning system, JCBP, groups the set of cases sharing the same goal in each domain into individual case bases and maintains indexes to these individual case bases. The system applies the heuristic knowledge automatically extracted from the problem model to the case adaptation phase. It provides a sort of case generalization through goal regression. Also JCBP can operate in an interactive mode to support a mixed-initiative planning. Since it considers and utilizes user's preference and knowledge for solving the given planning problems, it can generate solution plans satisfying more user's needs and reduce the complexity of plan generation.