• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.10
• /
• pp.4092-4107
• /
• 2015

We propose a background subtraction method for moving cameras based on trajectory classification, image segmentation and label inference. In the trajectory classification process, PCA-based outlier detection strategy is used to remove the outliers in the foreground trajectories. Combining optical flow trajectory with watershed algorithm, we propose a trajectory-controlled watershed segmentation algorithm which effectively improves the edge-preserving performance and prevents the over-smooth problem. Finally, label inference based on Markov Random field is conducted for labeling the unlabeled pixels. Experimental results on the motionseg database demonstrate the promising performance of the proposed approach compared with other competing methods.

• Journal of KIISE
• /
• v.42 no.12
• /
• pp.1561-1567
• /
• 2015

Abstract Semi-supervised learning is an area in machine learning that employs both labeled and unlabeled data in order to train a model and has the potential to improve prediction performance compared to supervised learning. Graph-based semi-supervised learning has recently come into focus with two phases: graph construction, which converts the input data into a graph, and label inference, which predicts the appropriate labels for unlabeled data using the constructed graph. The inference is based on the smoothness assumption feature of semi-supervised learning. In this study, we propose an enhanced label inference algorithm by incorporating the importance of each vertex. In addition, we prove the convergence of the suggested algorithm and verify its excellence.

• Child Studies in Asia-Pacific Contexts
• /
• v.5 no.2
• /
• pp.83-94
• /
• 2015

The purpose of this study was to understand the influence of personality trait information on young children's perception of initial attraction in peer relationships. The sample consisted of 90 children of three to five years of age in South Korea. Children were presented with an inductive inference task where they had to make inference of a target character's preference on novel-play and prosocial act based on trait labels (smart-not smart, outgoing-shy, nice-mean) and perceptual (toy) similarity information of two test characters. Children showed difference in their use of trait information depending on the perceptual similarity information, trait valence, and inference question with age. This result provides initial support that not only do young children understand the significance of trait in peer attraction but also know when trait label is more informative to use to infer others depending on the situation.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.12
• /
• pp.77-84
• /
• 2022

In this paper, we propose a Label Reassignment Strategy to improve the performance of an object detection algorithm. Our approach involves two stages: an inference stage and an assignment stage. In the inference stage, we perform multi-scale inference with predefined scale sizes on a trained model and re-infer masked images to obtain robust classification results. In the assignment stage, we calculate the IoU between bounding boxes to remove duplicates. We also check box and class occurrence between the detection result and annotation label to re-assign the dominant class type. We trained the YOLOX-L model with the re-annotated dataset to validate our strategy. The model achieved a 3.9% improvement in mAP and 3x better performance on AP_S compared to the model trained with the original dataset. Our results demonstrate that the proposed Label Reassignment Strategy can effectively improve the performance of an object detection model.

• Journal of Intelligence and Information Systems
• /
• v.7 no.1
• /
• pp.161-175
• /
• 2001

In this paper, a detection method of contradiction between input informations is proposed when the inference is processed in rule-based systems. The proposed method is accomplished by improving the label representation and the label management scheme in a conventional ATMS(Assumption-based Truth Maintenance System). The Proposed method also can represent and process input informations having uncertainty values.

• Journal of Global Scholars of Marketing Science
• /
• v.9
• /
• pp.47-75
• /
• 2002

The purpose of this study is to identify the effects of reference price, price rationale and price-quality inference of consumer on the evaluation of apparel quality. The experimental materials developed for this study were a set of stimulus and response sheet. The stimuli were six print ads, which was manipulated by reference price and price rationale for a jacket of national brand. This study used a 2(reference price: offer and non offer)$\times$3(price rationale: non offer, stock disposal, sales promotion) $\times$2(price-quality inference of consumer: high and low level) between-subjects experiment. Subjects were 371 female university students. The data were analyzed by factor analysis, ANOVA and t-test. The results were as follows. First, three apparel attributes were identified: sewing/fabrics and label by factor analysis. Second, the significant interaction effects of reference price, price rationale and price-quality inference of consumer were found on evaluating quality of sewing/fabrics and label of apparel. So, reference price effect differed to depending on type of price rationale and levels of price-quality inference. Third, the significant main effect of price-quality inference of consumer existed on evaluating construction quality of apparel.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2002.04a
• /
• pp.274-283
• /
• 2002

This paper presents an efficient models for reinforeced concrete structures using CART-ANFIS(classification and regression tree-adaptive neuro fuzzy inference system). a fuzzy decision tree parttitions the input space of a data set into mutually exclusive regions, each of which is assigned a label, a value, or an action to characterize its data points. Fuzzy decision trees used for classification problems are often called fuzzy classification trees, and each terminal node contains a label that indicates the predicted class of a given feature vector. In the same vein, decision trees used for regression problems are often called fuzzy regression trees, and the terminal node labels may be constants or equations that specify the Predicted output value of a given input vector. Note that CART can select relevant inputs and do tree partitioning of the input space, while ANFIS refines the regression and makes it everywhere continuous and smooth. Thus it can be seen that CART and ANFIS are complementary and their combination constitutes a solid approach to fuzzy modeling.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2001.06a
• /
• pp.377-381
• /
• 2001

An ATMS(Assumption-based Truth Maintenance System) has been widely used for maintaining the truth of information by detecting and solving contradictions in nile-based systems. But the ATMS can not correctly maintain the truth of the information in case that the generated information is satisfied within a time interval or includes data about temporal relations of events in time varying situations, because it has no mechanism manipulating temporal data. In this paper, The extended ATMS method is proposed, which can maintain the truth of the information in the inference system using information changing over time or temporal relations of events. In order to maintain contexts generated by relations of events, the label representation method is modified, the disjunction, conjunction simplification method in the label-propagation procedure and nogood handling method of the conventional ATMS are modified, too.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.352-357
• /
• 2023

자연어 추론은 두 문장(전제, 가설)간의 관계를 이해하고 추론하여 함의, 모순, 중립 세 가지 범주로 분류하며, 전제-가설-라벨(PHL) 데이터셋을 활용하여 자연어 추론 모델을 학습한다. 그러나, 새로운 도메인에 자연어 추론을 적용할 경우 학습 데이터가 존재하지 않거나 이를 구축하는 데 많은 시간과 자원이 필요하다는 문제가 있다. 본 논문에서는 자연어 추론을 위한 학습 데이터인 전제-가설-라벨 삼중항을 자동 생성하기 위해 [1]에서 제안한 문장 변환 규칙 대신에 거대 언어 모델과 Chain-of-Thought(CoT), Program-aided Language Models(PaL) 등의 프롬프팅(Prompting) 방법을 이용하여 전제-가설-라벨 삼중항을 자동으로 생성하는 방법을 제안한다. 실험 결과, CoT와 PaL 프롬프팅 방법으로 자동 생성된 데이터의 품질이 기존 규칙이나 기본 프롬프팅 방법보다 더 우수하였다.

• Annual Conference on Human and Language Technology
• /
• 2022.10a
• /
• pp.170-175
• /
• 2022

일반적으로 대규모 언어 모델들은 다량의 데이터를 오랜시간 사전학습하면서 레이블을 예측하기 위한 성능을 높여왔다. 최근 언어 모델의 레이블 예측에 대한 정확도가 높아지면서, 언어 모델이 왜 해당 결정을 내렸는지 이해하기 위한 신뢰도 높은 Natural Language Explanation(NLE) 을 생성하는 것이 시간이 지남에 따라 주요 요소로 자리잡고 있다. 본 논문에서는 높은 레이블 정확도를 유지하면서 동시에 언어 모델의 예측에 대한 신뢰도 높은 explanation 을 생성하는 참신한 자연어 추론 시스템을 제시한 Natural-language Inference over Label-specific Explanations(NILE)[1] 을 소개하고 한국어 데이터셋을 이용해 NILE 과 NLE 를 활용하지 않는 일반적인 자연어 추론 태스크의 성능을 비교한다.