• Proceedings of the Korean Society of Computer Information Conference
• /
• 한국컴퓨터정보학회 2016년도 제54차 하계학술대회논문집 24권2호
• /
• pp.51-54
• /
• 2016

In computer vision especially in image processing, it has become popular to apply deep convolutional networks for supervised learning. Convolutional networks have shown a state of the art results in classification, object recognition, detection as well as semantic segmentation. However, supervised learning has two major disadvantages. One is it requires huge amount of labeled data to get high accuracy, the second one is to train so much data takes quite a bit long time. On the other hand, unsupervised learning can handle these problems more cheaper way. In this paper we show efficient way to learn features for classification in an unsupervised way. The network trained layer-wise, used backpropagation and our network learns features from unlabeled data. Our approach shows better results on Caltech-256 and STL-10 dataset.

• KOREAN JOURNAL OF CROP SCIENCE
• /
• 제34권s01호
• /
• pp.40-47
• /
• 1989

In spite of the development of highly sophisticated instrument, the precise quantitation of plant hormones still has many difficulties. Due to their high specificity, sensitivity and minimal sample purification steps, immunological assays have been widely applied for plant hormone assay. Enzme-linked immunosorbent assay technique for the determination of plant hormones was developed by Voller in 1978. Immunological assays are accomplished by competition of labeled tracer antigen and unlabeled antigen for a limited number of specific antibodies. The use of enzyme as replacement labels for radioisotopes enabled much of the sensitivity and specificity of radioimmunoassay (RIA) to be retained but without the inherent disadvantage of high capital cost, potential health hazard, and short shelf life of the labeled reactants.

• Journal of Plant Biology
• /
• 제37권2호
• /
• pp.151-158
• /
• 1994

Aleurone layers of normal and vp1 mutant maize kernels were extracted and centrifuged at 100,000g to yield a cytosol fraction. Binding of [3H]ABA cis, trans (＋)ABA to a soluble macromolecular components present in the cytosol was demonstrated by Sephadex chromatography and non-denaturing PAGE. The binding component was of high molecular weight and seems to be an aggregate of proteins. A rapid DEAE-cellulose filter method for assaying bound [3H]ABA to a soluble protein was adapted. Binding assays were performed with cytosol that had been preheated or incubated with several enzymes, indicating that heat and protease treatments disrupted the binding. This suggested that binding occurred to proteins. Some properties of the ABA binding proteins were described. The [3H]ABA binding were reduced dramatically when unlabeled ABA was added as a competitor, suggesting a specific binding of [3H]ABA. Gel filtration profiles and autoradiogram of [3H]ABA binding showed no difference in the binding components of Vp1 and vp1/vp1 mutant cytosol, indicating that Vp1 protein is not a sole ABA binding protein.

• ETRI Journal
• /
• 제32권5호
• /
• pp.766-773
• /
• 2010

Support vector machine (SVM) active learning plays a key role in the interactive content-based image retrieval (CBIR) community. However, the regular SVM active learning is challenged by what we call "the small example problem" and "the asymmetric distribution problem." This paper attempts to integrate the merits of semi-supervised learning, ensemble learning, and active learning into the interactive CBIR. Concretely, unlabeled images are exploited to facilitate boosting by helping augment the diversity among base SVM classifiers, and then the learned ensemble model is used to identify the most informative images for active learning. In particular, a bias-weighting mechanism is developed to guide the ensemble model to pay more attention on positive images than negative images. Experiments on 5000 Corel images show that the proposed method yields better retrieval performance by an amount of 0.16 in mean average precision compared to regular SVM active learning, which is more effective than some existing improved variants of SVM active learning.

• Proceedings of the Korea Information Processing Society Conference
• /
• 한국정보처리학회 2003년도 추계학술발표논문집 (상)
• /
• pp.283-286
• /
• 2003

군집화 시 일반적으로 개별 클래스(class) 혹은 카테고리(category) 당 하나의 군집이 형성되는 결과가 선호된다. 하지만 데이터가 비정형적인 분포를 따르는 경우에는 하나의 군집으로 개별 클래스를 온전히 표현하는 것이 불가능하거나 오히려 부자연스러운 경우가 발생할 수 있다. 본 논문에서는 예제의 클래스를 알고 있는 즉, 레이블(label)된 예제들을 그렇지 않은(unlabeled) 예제들과 함께 활용하여 군집화하는 제약된 K-means (constrained K-means) 알고리즘을 위하여 보다 자연스러운 형태의 군집이 형성될 수 있도록 초기 씨드(seed, 씨앗)를 생성하는 방안을 제안한다. 레이블된 예제들을 계층적으로 군집화하면 다양한 단계에서 제약된 K-means를 위한 씨드집합을 생성할 수 있다. 본 연구에서는 각 단계의 씨드집합을 기반으로 형성된 군집결과간의 변화정도를 측정하여 가장 적절한 것으로 추정되는 씨드집합을 선정하였다. 제안한 방안을 문서 군집화 문제에 적용하여 실험한 결과 개별 클래스마다 하나의 군집을 가정하는 경우보다 더 나은 군집을 형성할 수 있음을 확인하였다.

• Annual Conference on Human and Language Technology
• /
• 한국정보과학회언어공학연구회 2018년도 제30회 한글 및 한국어 정보처리 학술대회
• /
• pp.608-610
• /
• 2018

한국어 의존 구문 분석(Dependency Parsing)은 문장 어절의 중심어(head)와 수식어(modifier)의 의존관계를 표현하는 자연어 분석 방법이다. 최근에는 이러한 의존 관계를 표현하기 위해 주의 집중 메커니즘(Attention Mechanism)과 LSTM(Long Short Term Memory)을 결합한 모델들이 높은 성능을 보이고 있다. 본 논문에서는 개선된 Biaffine Attention 의존 구문 분석 모델을 제안한다. 제안된 모델은 기존의 Biaffine Attention에서 의존성과 의존 관계를 결정하는 방법을 개선하였고, 한국어 의존 구문 분석을 위한 입력 열의 형태소 표상을 확장함으로써 기존의 모델보다 UAS(Unlabeled Attachment Score)가 0.15%p 더 높은 성능을 보였다.

• Journal of the Korean Data and Information Science Society
• /
• 제25권2호
• /
• pp.455-464
• /
• 2014

Unlabeled examples are easier and less expensive to obtain than labeled examples. Semisupervised approaches are used to utilize such examples in an eort to boost the predictive performance. This paper proposes a novel semisupervised classication method named transductive least squares support vector machine (TLS-SVM), which is based on the least squares support vector machine. The proposed method utilizes the dierence convex algorithm to derive nonconvex minimization solutions for the TLS-SVM. A generalized cross validation method is also developed to choose the hyperparameters that aect the performance of the TLS-SVM. The experimental results conrm the successful performance of the proposed TLS-SVM.

• Biomolecules & Therapeutics
• /
• 제14권1호
• /
• pp.45-49
• /
• 2006

The purpose of this study was to clarify the efflux transport system of choline from brain to blood across the blood-brain barrier (BBB) in rats using the brain efflux index (BEI) method. $[^3H]$Choline was micro-injected into parietal cortex area 2 (Par2) of the rat brain, and was eliminated from the brain with elimination halflife of 45 min. The BBB efflux clearance of $[^3H]$choline was about 124 mL/min/g brain, which was determined from combination of an elimination rate constant $(1.54X10^{-2}min^{-1})$ and the distribution volume in the brain (8.05 mL/g brain). The efflux of $[^3H]$choline was inhibited by unlabeled choline in a dose-dependent manner and was significantly inhibited by cationic substrates, such as hemicholinium-3 and tetraethylammonium (TEA). These results suggest that the BBB may act as an efflux pump for choline to reduce the excessive choline concentration in the brain interstitial fluid.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권8호
• /
• pp.3790-3803
• /
• 2018

Content-based image retrieval is an approach used to query images based on their semantics. Semantic based retrieval has its application in all fields including medicine, space, computing etc. Semantically generated binary hash codes can improve content-based image retrieval. These semantic labels / binary hash codes can be generated from unlabeled data using convolutional autoencoders. Proposed approach uses semi-supervised deep hashing with semantic learning and binary code generation by minimizing the objective function. Convolutional autoencoders are basis to extract semantic features due to its property of image generation from low level semantic representations. These representations of images are more effective than simple feature extraction and can preserve better semantic information. Proposed activation and loss functions helped to minimize classification error and produce better hash codes. Most widely used datasets have been used for verification of this approach that outperforms the existing methods.

• Journal of Microbiology
• /
• 제35권1호
• /
• pp.47-52
• /
• 1997

To examine whether the xylene component of BTX (benzene, toluene, xylene) mixture is cometabolized and residues are produced in soil, $\^$14/C-labeled-0-xylene was added to sandy loam in combination with unlabeled benzene and toluene. After 4 weeks of incubation in a sealed system connected to an oxygen reservoir, 55.1% of the radiocarbon was converted to $\^$14/CO$\sub$2/, 3.0% was to 95.8% radiocarbon recovery. Biomass incorporation of o-xylene radiocarbon which was detected by fumigation/extraction was usually low (5.6%), but 32.1% radiocarbon became associated with soil humus. Most of the numus-bound radiocarbon was found in humin fraction. In addition to o-xylene, p-xylene and toluene also showed similar results. The evidence shows that some of their reactive methylcatechol biodegradation intermediates attach to the humic metrix in soil in preference to mineralization and biomass incorporation.