Earticle

본 연구의 목적은 실험수업에 참여한 세 명의 중학교 1학년 학생들의 모델 정교화 과정을 질적 사례연구를 통하여 살펴보고 모델 정교화에 영향을 미치는 요인을 분석하여 이에 관한 정보를 제공하고자 함이다. 분석 결과 학생들의 수학적 모델링의 정교화 과정은 각 단계가 단선적으로 일어나는 것이 아니라 모델 내에서 해를 구할 수 없거나 구한 해가 의미가 없는 경우 실제 문제 구성 단계로 돌아가 실제 문제를 수정하거나 수학적 모델 구성 단계로 돌아 가 모델을 수정 또는 정교화하는 것을 알 수 있었으며, 모델 정교화 과정에 영향을 미치는 요인으로는 모델링 문제, 메타인지적 사고, 교사․동료와의 의사소통 및 교사의 역할 등이 복합적으로 작용함이 나타났다.

The present qualitative case study explored the ways in which three middle school students constructed and refined their mathematical models and modeling processes, and factors that had influenced such refinement. The results suggest that students’ modeling processes are non-sequential in that the participant students reformulated their initial problem from the real-world problem situation and revised the model when they could not get a satisfactory solution or the acquired solution did not make sense. Moreover, the students’ model refinement processes were affected by the following four elements: the types of real-word problem situations, students’ metacognitive thinking, communications between teachers and peers, and the role of teachers.

We present a novel method aimed at refining ground truth data through regularization and modification, particularly applicable when working with the original ground truth set. Enhancing the performance of deep neural networks is achieved by applying regularization techniques to the existing ground truth data. In many machine learning tasks requiring pixel-level segmentation sets, accurately delineating objects is vital. However, it proves challenging for thin and elongated objects such as blood vessels in X-ray coronary angiography, often resulting in inconsistent generation of ground truth data. This method involves an analysis of the quality of training set pairs - comprising images and ground truth data - to automatically regulate and modify the boundaries of ground truth segmentation. Employing the active contour model and a recursive ground truth generation approach results in stable and precisely defined boundary contours. Following the regularization and adjustment of the ground truth set, there is a substantial improvement in the performance of deep neural networks.

The stiffness of a structure is one of several structural signals that are useful indicators of the amount of damage that has been done to the structure. To accurately estimate the stiffness, an equation of motion containing a stiffness parameter must first be established by expansion as a linear series model, a Taylor series model, or a power series model. The model is then used in multivariate autoregressive modeling to estimate the structural stiffness and compare it to the theoretical value. Stiffness assessment for modeling purposes typically involves the use of one of three statistical model refinement approaches, one of which is the efficient Akaike information criterion (AIC) proposed in this paper. If a newly added component of the model results in a decrease in the AIC value, compared to the value obtained with the previously added component(s), it is statistically justifiable to retain this new component; otherwise, it should be removed. This model refinement process is repeated until all of the components of the model are shown to be statistically justifiable. In this study, this model refinement approach was compared with the two other commonly used refinement approaches: principal component analysis (PCA) and principal component regression (PCR) combined with the AIC. The results indicate that the proposed AIC approach produces more accurate structural stiffness estimates than the other two approaches.

A new model of TEE (Teaching English in English) teaching skill development is presented. Grounded in cognitive acquisition and information process, The dedicated cognitive process (DCP) model provides a comprehensive framework that account for a range of phenomena encountered by trainee teachers in the TEE program - for example, why different training methods are needed for the different elements of teaching skill. The model features three principal systems: declarative knowledge, procedural knowledge and the reflective system. The reflection system is identified as central to teachers' TEE skill development and, accordingly, a pivotal role is given to a reflective system, which enables teachers to reflect and build on both their declarative knowledge and procedural knowledge. The DCP model incorporates teachers skills and explains why different skills should be developed in different ways. It highlights the centrality of teachers linguistic ability, classroom plan, procedures and skills in the development of teachers' expertise. It makes a distinction between personal and professional teachers' roles and it also identifies the role of personal relationship skills in the TEE program. While there are still many questions to be investigated, it is hoped that the model will stimulate more research projects and provide understanding and guidance for the TEE program.

Employment of a time consuming, sophisticated calculation using the all-atom force field and generalized-Born implicit solvent model (GBIS) for refinement of NMR structures has become practical through advances in computational methods and capacities. GBIS refinement improves the qualities of the resulting NMR structures with reduced computational times. However, the contribution of GBIS to NMR structures has not been sufficiently studied in a quantitative way. In this paper, we report the effects of GBIS on the refined NMR structures of ubiquitin (UBQ) and GB1 with subsets of distance restraints derived from experimental data. Random omission prepared a series of distance restraints 0.05, 0.1, 0.3, 0.5, and 0.7 times smaller. For each number, we produced five different restraints for statistical analysis. We then recalculated the NMR structures using CYANA software, followed by GBIS refinements using the AMBER package. GBIS improved both the precision and accuracy of all the structures, but to varied levels. The degrees of improvement were significant when the input restraints were insufficient. In particular, GBIS enabled GB1 to form an accurate structure even with distance restraints of 5%, revealing that the root-mean-square deviation was less than 1 ${\AA}$ from the X-ray backbone structure. We also showed that the efficiency of searching the conformational space was more important for finding accurate structures with the calculation of UBQ with 5% distance restraints than the number of conformations generated. Our data will provide a meaningful guideline to judge and compare the structural improvements by GBIS.

There are two distinct approaches to improving the quality of protein NMR structures during refinement: all-atom force fields and accumulated knowledge-assisted methods that include Rosetta. Mao et al. reported that, for 40 proteins, Rosetta increased the accuracies of their NMR-determined structures with respect to the X-ray crystal structures (Mao et al., J. Am. Chem. Soc. 136, 1893 (2014)). In this study, we calculated 32 structures of those studied by Mao et al. using all-atom force field and implicit solvent model, and we compared the results with those obtained from Rosetta. For a single protein, using only the experimental NOE-derived distances and backbone torsion angle restraints, 20 of the lowest energy structures were extracted as an ensemble from 100 generated structures. Restrained simulated annealing by molecular dynamics simulation searched conformational spaces with a total time step of 1-ns. The use of GPU-accelerated AMBER code allowed the calculations to be completed in hours using a single GPU computer-even for proteins larger than 20 kDa. Remarkably, statistical analyses indicated that the structures determined in this way showed overall higher accuracies to their X-ray structures compared to those refined by Rosetta (p-value < 0.01). Our data demonstrate the capability of sophisticated atomistic force fields in refining NMR structures, particularly when they are coupled with the latest GPU-based calculations. The straightforwardness of the protocol allows its use to be extended to all NMR structures.

Refinement of NMR structures by molecular dynamics (MD) simulations with a solvent model has improved the structural quality. In this study, we applied MD refinement with the generalized Born (GB) implicit solvent model to protein structure determined under membrane-like environments. Despite popularity of the GB model, its applications to the refinement of NMR structures of hydrophobic proteins, in which detergents or organic solvents enclose proteins, are limited, and there is little information on the use of another GB parameter for these cases. We carried out MD refinement of crambin NMR structure in dodecylphosphocholine (DPC) micelles (Ahn et al., J. Am. Chem. Soc. 2006, 128, 4398-4404) with GB/Surface area model and two different surface tension coefficients, one for aquatic and the other for hydrophobic conditions. Our data show that, of two structures by MD refinement with GB model, the one refined with the parameter to consider hydrophobic condition had the better qualities in terms of precision and solvent accessibility.

본 연구에서는 2차원 비압축성 유체 유동을 해석하기 위하여 격자 세밀화 모델을 적용한 격자볼츠만 방법(LBM)을 수치적으로 연구하였다. 일반적으로 유동해석에서 충분한 정확도를 얻기 위해서는 관심 있는 영역에서 격자가 세밀하게 구성되어야 한다. 그래서 본 연구에서는 유동함수-와도 공식을 적용한 LBM에 격자 세밀화 모델을 적용하여 유동해석을 수행하였다. 공동형상 유동에서의 기존의 신뢰성 있는 유동장 결과와의 비교를 통해 본 연구의 격자 세밀화 모델을 적용한 격자볼츠만 방법의 신뢰성과 유용성을 검토하였다.

In this study, we present a grid refinement model in the lattice Boltzmann method (LBM) for two-dimensional incompressible fluid flow. That is, the model combines the desirable features of the lattice Boltzmann method and stream function-vorticity formulations. In order to obtain an accurate result, very fine grid (or lattice) is required near the solid boundary. Therefore, the grid refinement model is used in the lattice Boltzmann method for stream function-vorticity formulation. This approach is more efficient in that it can obtain the same accurate solution as that in single-block approach even if few lattices are used for computation. In order to validate the grid refinement approach for the stream function-vorticity formulation, the numerical simulations of lid-driven cavity flows were performed and good results were obtained.

인간-컴퓨터 상호작용(HCI)에서 복잡한 감정 이해의 중요성이 커지면서 멀티모달 감정 인식 연구가 활발히 진행되고 있다. 최근에는 인간의 복잡한 감정을 인식하기 위해 다양한 모달리티를 사용하여 연구되고 있다. 그러나 기존 연구는 한국어 특유의 복합적인 감정 표현에서 나타나는 모달리티 간 상호작용을 충분히 포착하는 데 한계가 있다. 본 연구는 이를 극복하기 위해 텍스트, 음성, 자모 정보를 결합하는 순차적 정제 기반 멀티모달 융합 모델을 제안한다. 먼저 KLUE-BERT, HuBERT, 자모 기반 Transformer Encoder로 각 모달리티의 특징을 추출하고, Cross-Modal Attention을 통해 특징을 순차적으로 정제하며 융합한다. 이후 Conformer 블록과 Attention Pooling을 적용해 핵심 표현을 도출한다. 제안 모델은 정확도 0.8462를 달성하였다. 이는 모달리티 간 상관관계를 정밀하게 모델링하여 한국어 감정 인식 성능을 유의미하게 향상시킴을 보여준다.

As the importance of understanding complex emotions in human-computer interaction (HCI) has grown, research on multimodal emotion recognition has become increasingly active. Recently, diverse modalities have been leveraged to recognize nuanced human emotions. However, prior studies have limitations in sufficiently capturing inter-modal interactions that arise in Korean-specific, compound emotional expressions. To address this gap, we propose a sequential refinement-based multimodal fusion model that integrates text, speech, and jamo (Korean grapheme) information. Specifically, we extract modality-specific representations using KLUE-BERT, HuBERT, and a jamo-based Transformer encoder, and then sequentially refine and fuse these features via cross-modal attention. Subsequently, we apply Conformer blocks and attention pooling to derive salient expressive cues. The proposed model achieves an accuracy of 0.8462, demonstrating that precisely modeling cross-modal correlations can meaningfully improve Korean emotion recognition performance.

이 논문은 스프레드시트를 활용한 수학적 모델링에서 스프레드시트 환경이 수학적 모델의 정교화과정에 어떤 영향을 미치는 지를 고찰한 것이다. 좀 더 자세히 살피면 수학적 모델링에서 스프레드시트 모델의 활용은 학생이 분석 불가능한 수학적모델도 분석할 수 있도록 해줌으로써 모델을 단순화하지 않고, 대신 모델을 정교화 할 수 있는 기회를 제공하고 수학적 개념을 확장해 나갈 수 있음을 보였다. 또한 수학적 모델을 스프레드시트 모델로 변환하여, 수학적 모델로부터 수학적 결론을 얻는 단계를 거치지 않고도 실세계 상황을 해석하고 설명할 수 있는 기회를 제공할 수 있음을 보였다.

In mathematical modeling activity modeling process is usually an iterative process. When model can not be solved, the model needs to be simplified by treating some variables as constants, or by ignoring some variables. On the other hand, when the results from the model are not precise enough, the model needs to be refined by considering additional conditions. In this study we investigate the role of spreadsheet model in model refinement and modeling process. In detail, we observed that by using spreadsheet model students can solve model which can not be solved in paper-pencil environment. And so they need not go back to model simplification process but continue model refinement. By transforming mathematical model to spreadsheet model, the students can predict or explain the real word situations directly without passing the mathematical conclusions step in modeling process.

일반적으로 의사결정의 대상이 되는 현실 시스템은 매우 가변적 (variable)이며 때로는 많은 불확실성(uncertainty)이 포함된 상황에 놓일 수 있다. 이러한 문제의 처리를 위한 통계적 방법으로 유의수준이나 확신도, 민감도 분석 등이 사용된다. 본 논문에서는 먼저 근접관계 행렬에서 근접도를 구하는 방법으로 상대적 해밍거리와 max-min방법을 이용한 다음, 다중임계치를 사용하여 최적구간분할을 하는 방법을 제안한다. 결과적으로 max-min방법을 이용하여 다중임계치을 적용한 근접관계의 분류가 상대적 해밍거리로 근접도를 구하여 다중임계치를 구하는 방법보다 계산과정이 더 간단하고 명확하며 분할과정을 줄일 수 있고 최적의 의사결정에 효율적이라는 것을 알 수 있다.

This paper proposes a convolutional neural network model for distinguishing areas occupied by obstacles from a LiDAR image converted from a 3D point cloud. The channels of a LiDAR image used as input consist of the distances to 3D points, the reflectivities of 3D points, and the heights of 3D points from the ground. The proposed model uses a LiDAR image as an input and outputs a result of a segmented LiDAR image. The proposed model adopts refinement modules with skip connections to segment a LiDAR image. The refinement modules with skip connections in the proposed model make it possible to construct a complex structure with a small number of parameters than a convolutional neural network model with a linear structure. Using the proposed model, it is possible to distinguish areas in a LiDAR image occupied by obstacles such as vehicles, pedestrians, and bicyclists. The proposed model can be applied to recognize surrounding obstacles and to search for safe paths.

For gasoline engines, a three-way catalyst converter is being used for cleaning up to exhaust gas contamination. Precise air/fuel ration control is necessary to maximize the catalytic conversion efficiency. In transient condition, feed-forward air/fuel ratio control that estimates air mass inducted into the cylinder is performed. In this study, a fuel injection map that makes accurate air/fuel ratio control possible was constructed for the very same transient condition. In the same condition above, intake air model and fuel model were refined so that fuel injection values based on air mass flow through throttle valve and intake manifold pressure are equal to fuel injection values of the map.

For gasoline engines, a three-way catalyst converter that has the maximum efficiency at stoichiometric air/fuel ratio is being used to clean up the exhaust gas. So a precise air/fuel ratio control is necessary to maximize the catalytic conversion efficiency. For a transient condition, a feed-forward air/fuel ratio control method that estimates the air mass inducted into a cylinder is being used. In this study, optimal fuel injection data that makes an accurate air/fuel ratio control possible was constructed for the very same transient condition. Then, fuel model were refined so that fuel injection values based on air mass through a throttle valve and intake manifold pressure has the least difference from optimal fuel injection values for the same condition.

GRACE는 2002년 발사된 이래로 지하수, 빙하, 해수면의 변동에 의한 지구의 질량 재배치를 관측해오고 있다. GRACE 관측으로부터 추정된 지구 중력 모델은 기압보정을 거쳐 대기 질량이 제거된 지구 표면과 그 하부의 질량 변화를 나타낸다. 그러나 대기 총 질량은 지표면과 대기 사이의 물 교환에 의해 변한다. 그 결과 GRACE 중력 모델은 구면 조화 함수의 계수 degree 0, order 0 ($C_{00}$)에 해당하는 총 대기 질량 변화에 관련된 중력 스펙트럼을 가져야 한다. 주로 계절적인 시간 척도 안에서 변하는 수증기 때문에 $C_{00}$의 변화(${\delta}C_{00}$)는 특히 해수면의 계절적 변동과 북반구와 남반구 사이의 질량 균형에 매우 중요하다. 이 결과는 ${\delta}C_{00}$가 기후변동과 관련된 대륙 규모의 질량 변화 연구에 꼭 고려되어야 함을 뜻한다.

The Gravity Recovery and Climate Experiment (GRACE) has observed the Earth's mass redistribution mainly caused by the variations of groundwater, ice sheet, and sea level since its launch in April 2002. The global gravity model estimated by the GRACE observation is corrected by barometric pressure, and thus represents the change of Earth mass on the Earth's surface and below Earth's surface excluding air mass. However, the total air mass varies due to the water exchange between the Earth's surface and the atmosphere. As a result, the nominal GRACE gravity model should include the Earth's gravity spectrum associated with the total air mass variations, degree 0 and order 0 coefficients of spherical harmonics ($C_{00}$). Because the water vapor content varies mainly on a seasonal time scale, a change of $C_{00}$ (${\delta}C_{00}$) is particularly important to seasonal variations of sea level, and mass balance between northern and southern hemisphere. This result implies that ${\delta}C_{00}$ coefficients should be accounted for the examination of continental scale mass change possibly associated with the climate variations.

As film products are increasingly used in a wide range of areas, from producing traditional flexible packaging to high-tech electronic products, a higher level of quality is demanded. Most film products are made in the form of rolled finished goods, therefore, various quality issues related to their shape characteristics must be addressed. The thickness of the film products is one of the most common and important critical-to-quality attributes (CTQs). Particularly, the degree of thickness uniformity is more important than other thickness parameters, because it will be potential causes of many secondary thickness-related quality problems, such as wrinkles or faulty windings. To control the degree of thickness uniformity, the fixed bending region is oneof the most important CTQs to manage. Fixed bending regions are special points in the transverse direction of a rolled product with consistent minute variations of the thickness gap. This paper describes the measurement and analysis of thickness uniformity data, which were performed in a real manufacturing field of biaxial oriented polypropylene (BOPP) film. In previous researches, quality function deployment (QFD) or fault tree analysis were used to find the most critical process attributes out to controlthe CTQ of thickness uniformity. Whereas, this paper uses traditional control charts to find the most critical process attributes out in this problem. In addition, the selection of one of the major critical process attributes (CTPs) that is expected to affect the CTQ of thickness uniformity is also described. The selected critical-to-process attributes are the controlled temperatures along the transverse direction. A dramatic improvement in thickness uniformity was observed when the selected CTPs were controlled.

Conceptual site model (CSM) development and enhancement for contaminated sites assists in identifying data gaps during the site investigation process. In this study, CSM was developed and enhanced for a contaminated site in Korea as a case study. Site Y was scrutinized four times previously. The site profiles for each scrutiny were reorganized based on the scrutiny reports, and the relevant data was utilized to develop and enhance CSMs. CSM for the first investigation was developed in various forms including table, flowchart, diagram, and narrative formats. CSM was enhanced in a stepwise manner by incorporating the updated profile information obtained in next investigation to existing CSM. The hypothetical data gap analysis between each investigation step was established to meet the purpose of the follow-up investigation. This case study showed that CSM is a useful tool to identify the history and current status of contaminated sites and thereby help in planning supplementary investigations for better site characterization.

자연어 기반의 분류모델을 개발할 때 높은 성능을 획득하기 위해서는 데이터의 품질이 중요한 요소이다. 특히 무역상품 국제 분류체계 HS-CODE에서 상품명을 기반으로 HS코드를 분류할 때, 라벨링 된 데이터의 품질에 의해서 분류모델의 성능이 좌우된다. 하지만 현실적으로 확보 가능한 데이터셋에는 데이터 라벨링 오류나 데이터로 활용하기에 특징점이 부족한 데이터들이 다수 존재하기도 한다. 본 연구에서는 분류모델 학습 데이터의 정제 방법론으로, 딥러닝 기반 노이즈 검출 알고리즘을 제안한다. 분류 대상의 특징점이 분류 경계값 주변에 존재한다면 분류하기 모호한 노이즈 데이터일 가능성이 높다고 가정하고, 해당 노이즈 데이터를 검출하는 방법으로 딥러닝 기술을 활용한다. 해당 경계값 노이즈 검출 알고리즘으로 데이터를 정제한 뒤 학습모델의 성능비교 결과, 기존 대비 우수한 분류 정확도를 기록하였다.

협업으로 문제를 해결할 때 집단 구성원들은 정신모형을 공유함으로써 시너지 효과를 창출한다. 협업의 시너지 효과를 이해하고 이를 교육에 적용하기 위해서는, 집단이 어떻게 문제를 해결해나가는지를 분석할 필요가 있다. 이 연구에서는 정신모형 공유에서 나타나는 시너지 효과의 원천을 알아보기 위하여 10명의 과학영재들을 대상으로 빈약하게 정의된 문제를 주고 어떤 요소가 집단의 정신모형에 공유되는지, 또 이것이 어떤 과정을 거쳐 공유되어 가는지를 구체적으로 알아보았다. 연구결과, 집단의 정신모형에 공유되는 요소는 첫째, 초기 공통 아이디어 중 가장 과학적인 개념을 공유하였다. 논의 전 이미 가지고 있던 공통 아이디어에 포함된 개념에서 선개념과 과학적 개념이 동시에 나타날 때, 학생들은 논의를 거쳐 과학적인 개념을 선택하여 공유하였다. 둘째, 집단 구성원들이 각각 가지고 있던 개별 아이디어를 공유하였다. 논의 과정에서 구성원들은 변인에 대한 개별 아이디어를 제안하였고 집단 구성원들이 동의하면 이는 집단의 정신모형에 공유되었다. 셋째, 협업 전 정신모형에 포함되어 있지 않던 새로운 아이디어를 창조하여 공유하였다. 건설적 갈등 상황에서 자신의 정당성 주장의 근거 마련을 위해 개별 아이디어 조합, 기존 아이디어의 재생산 등과 같은 방법으로 창조된 아이디어를 생성하고 공유하였다. 창조된 아이디어는 개인의 정신모형에 포함되어 있지 않던 것으로, 집단 정신모형 공유에서 나타나는 시너지의 근원이라고 할 수 있다. 다음으로 집단 정신모형이 공유되는 과정을 살펴보면, 문제해결에 가장 중요한 변인부터 제안하고 공유하는 것을 알 수 있었다. 이는 집단의 정신모형에 무엇이 공유되었는가를 분석하는 것만큼 어떤 순서로 공유되었는가를 아는 것도 중요하다는 점을 시사한다. 논의 후 학생들의 최종응답을 분석해보면, 공유된 정신모형이 초기의 개인별 정신모형에 비해 포괄적이면서 정교화 되었으며 각 개인의 정신모형도 확장되었음을 알 수 있었다. 이 연구 결과는 공유된 정신모형의 측정과 협력 학습모형 개발에 사용될 수 있을 것이다.

To understand the synergy of collaboration and to apply this understanding to education, an analysis of how a team solves a problem and the sharing of their mental models is needed. This paper analyzed two things qualitatively to find out the source of synergy in a collaborative problem-solving process. First, the sharing contents in team mental model and second, the process of sharing the team mental model. Ten gifted middle school students collaborated to solve an ill-defined problem called sunshine through foliage problem. The gifted students shared the following results after the collaboration: First, scientific concept prior to common idea or the idea that all group members have before the discussions; second, unique individual ideas of group members; and third, created ideas that were not originally in the personal mental model. With created ideas, the team model becomes more than the sum of individuals. According to the results of process analysis, in the process of sharing mental model, the students proposed and shared the most important variable first. This result implied that the analysis of the order of sharing ideas is important as much as finding shared ideas. Also, the result shows that through their collaboration, the gifted students' shared mental model became more refined and expanded as compared to their individual prior mental models. It is recommended that these results can be used to measure shared mental model and develop collaborative learning models for students.