• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.04a
• /
• pp.924-927
• /
• 2017

통계 모형에 기반을 둔 구문분석기는 자료 부족 문제에 취약하거나 장거리 의존관계와 같은 특정 언어현상에 대한 처리가 어렵다는 단점이 있다. 이러한 한계점을 극복하고자 본 연구진은 규칙에 기반을 둔 한국어 구문분석기를 개발하고 있다. 다른 구문 분석기와 다르게 형태소 단위 구문분석을 시도하며 생성 가능한 모든 구문분석 후보를 보여주는 것이 특징이다. 본 연구진의 기존 연구에서 개발한 한국어 구문분석기는 형태소의 입력순서와 구문분석 후보의 생성 순서에 의존하여 구문분석 후보를 순서화하였다. 그러나 생성되는 구문분석 후보 중 가장 정답에 가까운 구문분석 후보의 순위를 낮추기 위해서는 각 구문분석 트리가 특정한 점수를 가질 필요가 있다. 본 논문에서는 품사 태거(tagger)에서 출력하는 어절별 형태소의 순위에 따른 가중치, 수식 거리에 따른 가중치, 특정한 지배-의존 관계에 대한 가중치를 이용해 가중치 합을 가지는 구문분석 후보를 구성하고 이를 정렬하여 이전 연구보다 향상된 성능을 가진 한국어 구문분석기 모델을 제안한다. 실험은 본 연구진이 직접 구축한 평가데이터를 기반으로 진행하였으며 기존의 Unlabeled Attachment Score(UAS) 87.86%에서 제안 모델의 UAS 93.34%로 약 5.48의 성능향상을 확인할 수 있었다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.11
• /
• pp.5427-5445
• /
• 2019

While deep neural networks have achieved remarkable performance in representation learning, a huge amount of labeled training data are usually required by supervised deep models such as convolutional neural networks. In this paper, we propose a new representation learning method, namely generative adversarial networks (GAN) based bagging deep convolutional autoencoders (GAN-BDCAE), which can map data to diverse hierarchical representations in an unsupervised fashion. To boost the size of training data, to train deep model and to aggregate diverse learning machines are the three principal avenues towards increasing the capabilities of representation learning of neural networks. We focus on combining those three techniques. To this aim, we adopt GAN for realistic unlabeled sample generation and bagging deep convolutional autoencoders (BDCAE) for robust feature learning. The proposed method improves the discriminative ability of learned feature embedding for solving subsequent pattern recognition problems. We evaluate our approach on three standard benchmarks and demonstrate the superiority of the proposed method compared to traditional unsupervised learning methods.

• Journal of the Korean Magnetic Resonance Society
• /
• v.15 no.2
• /
• pp.115-127
• /
• 2011

A newly developed cost-effective approach to prepare $^{15}N/^{13}C$-labeled protein for NMR studies is presented. This method has been successfully applied to isotopically labeling of PTK6 SH2 domain and MTH 1880 protein. The production method generates cell density using a growing media containing $^{15}NH_4Cl$, $^{12}C_6$-D-glucose. Following a doubling time period for unlabeled metabolite exhaustion and then addition $^{13}C_6$-D-glucose into a M9 growing media, the cells are induced. Our results demonstrate that in order to get full incorporation of $^{13}C$, the isotopes are not totally required during the initial growth phase before induction. The addition of small amounts of $^{13}C_6$-D-glucose to the induction phase is sufficient to obtain more than 95% incorporation of isotopes into the protein. Our optimized protocol is two-thirds less costly than the classical method using $^{13}C$ isotope during the entire growth phase.

• Journal of information and communication convergence engineering
• /
• v.15 no.1
• /
• pp.43-48
• /
• 2017

Movie ratings are crucial for recommendation engines that track the behavior of all users and utilize the information to suggest items the users might like. It is intuitively appealing that information about the viewing preferences in terms of movie genres is sufficient for predicting a genre of an unlabeled movie. In order to predict movie genres, we treat ratings as a feature vector, apply a Bernoulli event model to estimate the likelihood of a movie being assigned a certain genre, and evaluate the posterior probability of the genre of a given movie by using the Bayes rule. The goal of the proposed technique is to efficiently use movie ratings for the task of predicting movie genres. In our approach, we attempted to answer the question: "Given the set of users who watched a movie, is it possible to predict the genre of a movie on the basis of its ratings?" The simulation results with MovieLens 1M data demonstrated the efficiency and accuracy of the proposed technique, achieving an 83.8% prediction rate for exact prediction and 84.8% when including correlated genres.

• Journal of Radiopharmaceuticals and Molecular Probes
• /
• v.6 no.1
• /
• pp.53-58
• /
• 2020

Pretargeting is a two-component strategy often used for tumor targeting to enhance the tumor-to-background ratio in cancer diagnosis as well as therapy. In the multistep strategy, the highly specific unlabeled monoclonal antibodies (mAbs) with the reactive site is allowed to get localized at tumor site first, and then small and fastclearing radiolabeled chelator with counter reactive site is administered which covalently attaches to mAbs via inverse electron demand Diels-Alder reaction (IEDDA). The catalyst-free IEDDA cycloaddition reaction between 1,2,4,5-tetrazines and strained alkene dienophiles aid with properties like selective bioconjugation, swift and high yielding bioorthogonal reactions are emergent in the development of radiopharmaceutical. Due to its fast pharmacokinetics, the in vivo formed radioimmunoconjugates can be imaged at earlier time points by short-lived radionuclides like ¹⁸F and ⁶⁸Ga; it can also reduce radiation damage to the normal cells. Ultimately, this review elucidates the updated status of pretargeting based on antibodies and IEDDA for tumor diagnosis (PET and SPECT) and therapy.

• BMB Reports
• /
• v.31 no.1
• /
• pp.48-52
• /
• 1998

$p16^{INK4}$, which is a 16-kDa polypeptide protein, inhibits the catalytic activity of the CDK4-cyclinD complex to suppress rumor growth. Both unlabeled and isotope-labeled human tumor suppressor $p16^{INK4}$ protein were overexpressed and purified to characterize biochemical and structural properties. The purified p16 binds to monomeric GST-CDK4 and exists in a monomer conformation for several weeks at $4^{\circ}C$. The circular dichroism (CD) data indicates that p16 contains high percentage of ${\alpha}$-helix and that the helix percentage maximized at pH value of 7.0. One-and two-dimensional nuclear magnetic resonance (NMR) data suggest that purified p16 from our construct has a unique folded conformation under our experimental conditions and exhibits quite stable conformational characteristics.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.13 no.3
• /
• pp.186-199
• /
• 2013

We pose pattern classification as a density estimation problem where we consider mixtures of generative models under partially labeled data setups. Unlike traditional approaches that estimate density everywhere in data space, we focus on the density along the decision boundary that can yield more discriminative models with superior classification performance. We extend our earlier work on the recursive estimation method for discriminative mixture models to semi-supervised learning setups where some of the data points lack class labels. Our model exploits the mixture structure in the functional gradient framework: it searches for the base mixture component model in a greedy fashion, maximizing the conditional class likelihoods for the labeled data and at the same time minimizing the uncertainty of class label prediction for unlabeled data points. The objective can be effectively imposed as individual mixture component learning on weighted data, hence our mixture learning typically becomes highly efficient for popular base generative models like Gaussians or hidden Markov models. Moreover, apart from the expectation-maximization algorithm, the proposed recursive estimation has several advantages including the lack of need for a pre-determined mixture order and robustness to the choice of initial parameters. We demonstrate the benefits of the proposed approach on a comprehensive set of evaluations consisting of diverse time-series classification problems in semi-supervised scenarios.

• Journal of the Korea Society of Computer and Information
• /
• v.22 no.7
• /
• pp.17-29
• /
• 2017

IRecently, researches on the recognition of indoor user situations through various sensors in a smart home environment are under way. In this paper, the case study was conducted to determine the operation of the robot vacuum cleaner by inferring the user 's indoor situation through the operation of home appliances, because the indoor situation greatly affects the operation of home appliances. In order to collect learning data for indoor situation awareness model learning, we received feedbacks from user when there was a mistake about the cleaning situation. In this paper, we propose a semi-supervised learning method using user feedback data. When we receive a user feedback, we search for the labels of unlabeled data that most fit the feedbacks collected through genetic algorithm, and use this data to learn the model. In order to verify the performance of the proposed algorithm, we performed a comparison experiments with other learning algorithms in the same environment and confirmed that the performance of the proposed algorithm is better than the other algorithms.

• Bulletin of the Korean Chemical Society
• /
• v.35 no.2
• /
• pp.383-391
• /
• 2014

Biotin-S-S-Phosphine was designed and synthesized as a potential tool for a proteomic study of O-GlcNAcmodified proteins. This reagent features a disulfide linker between a triarylphosphine moiety, which allows selective conjugation to azide-containing proteins, and a biotin moiety that can allow easy isolation through its strong affinity toward avidin-coated solid beads. The disulfide linkage within this reagent can allow the easy release of the bound molecules of interest, which is difficult to achieve when a biotin:avidin pair is used alone, by reducing the disulfide bond of the reagent with DTT. Preliminary in vitro biological assays with azidelabeled and unlabeled cell lysates and a pure protein Nup62 showed that the Biotin-S-S-Phosphine reagent is highly reactive toward the free thiol groups of proteins. When a molecular tool with a disulfide linker is applied to the enrichment of the molecules of interest from other species, it is important to block the free-thiols of the sample using exhaustive alkylation prior to the Staudinger ligation reactions to restore the bioorthogonal nature of this reaction.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.2
• /
• pp.165-171
• /
• 2017

It is difficult to recognize Human's facial expression in the real-world. For these reason, when database and test data have similar condition, we can accomplish high accuracy. Solving these problem, we need to many facial expression data. In this paper, we propose the algorithm for gathering many facial expression data within various environment and gaining high accuracy quickly. This algorithm is training initial model with the ASSL (Active Semi-Supervised Learning) using deep learning network, thereafter gathering unlabeled facial expression data and repeating this process. Through using the ASSL, we gain proper data and high accuracy with less labor force.