• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.5
• /
• pp.592-601
• /
• 2016

The recurrent neural network (RNN) is a deep learning model which is suitable to sequential or length-variable data. The Long Short-Term Memory (LSTM) mitigates the vanishing gradient problem of RNNs so that LSTM can maintain the long-term dependency among the constituents of the given input sequence. In this paper, we propose a LSTM based language model which can predict following words of a given incomplete sentence to generate a complete sentence. To evaluate our method, we trained our model using multiple Korean corpora then generated the incomplete part of Korean sentences. The result shows that our language model was able to generate the fluent Korean sentences. We also show that the word based model generated better sentences compared to the other settings.

• Journal of KIISE:Software and Applications
• /
• v.26 no.12
• /
• pp.1428-1444
• /
• 1999

MSC는 ITU에 의해 표준화된, 병행 시스템의 명세를 기술하기 위한 그래픽 형태와 텍스트 형태를 제공하는 언어로서 실시간 시스템 특히 통신 교환 시스템의 특성을 기술하기 위해 자주 사용된다. 통신 시스템이 제대로 동작함을 보이기 위해서는 정형적인 방법을 사용하여 시스템 행위를 검증할 필요가 있다. 통신 소프트웨어를 검증하는 방법 중 하나의 방법으로 유한 상태를 기반으로 하는 방법이 있다. 유한 상태를 기반으로 하는 방법에서는 먼저 시스템 명세에 해당하는 전체 상태 전이 그래프를 생성한 후 이를 바탕으로 model-checking 등의 방법을 사용하여 시스템의 특성을 검증한다. 본 논문에서는 MSC로 기술된 통신 소프트웨어 명세로부터 전체 상태 전이 그래프를 생성하는 방법에 초점을 맞추었다. 시스템 명세에 해당하는 상태 전이 그래프를 생성하기 위해 보다 직관적으로 MSC의 의미론을 표현할 수 있고, 또한 쉽게 전체 상태 전이 그래프를 생성할 수 있는 행위 종속 그래프를 제안하였다. MSC 명세는 일단 행위 종속 그래프로 변환이 되고 이 행위 종속 그래프를 이용하여 전체 상태 전이 그래프가 생성된다.Abstract Message Sequence Chart (MSC) standardized by International Telecommunication Union is a graphical and textual language for describing the specification of concurrent systems. It is frequently used both formally and informally for specifying the behavior of real-time systems, in particular telecommunication switching systems. To ensure that a communication system operates properly, the verification process showing the correctness of system's behavior formally is necessary. One of the verification methods is a finite-state method. In the finite-state method, the global state transition graph (GSTG) is constructed and then safety and liveness properties of systems are verified through a well-known method such as model checking. In this paper, we forcus on the construction of GSTG from the specifications of telecommunication software written in MSC. We suggest Action Dependency Graph (ADG) which can present the semantics of MSC intuitively and also provide a GSTG construction method from ADG. MSC specifications are translated to ADGs and, in turns, the GSTGs are constructed by using these ADGs.

• Proceedings of the KSRS Conference
• /
• 2009.03a
• /
• pp.97-100
• /
• 2009

The surface parameters associated with the land are usually dependent on the climate, and many physical processes that are displayed in the image sensed from the land then exhibit temporal variation with seasonal periodicity. An adaptive feedback system proposed in this study reconstructs a sequence of images remotely sensed from the land surface having the physical processes with seasonal periodicity. The harmonic model is used to track seasonal variation through time, and a Gibbs random field (GRF) is used to represent the spatial dependency of digital image processes. In this study, the Normalized Difference Vegetation Index (NDVI) was computed for one week composites of the Advanced Very High Resolution Radiometer (AVHRR) imagery over the Korean peninsula for 1996 and 2000 using a dynamic technique, and the adaptive reconstruction of harmonic model was then applied to the NDVI time series for tracking changes on the ground surface. The results show that the adaptive approach is potentially very effective for continuously monitoring changes on near-real time.

• Korean Journal of Remote Sensing
• /
• v.25 no.2
• /
• pp.95-105
• /
• 2009

Irregular temporal sampling is a common feature of geophysical and biological time series in remote sensing. This study proposes an on-line system for reconstructing observation image series including bad or missing observation that result from mechanical problems or sensing environmental condition. The surface parameters associated with the land are usually dependent on the climate, and many physical processes that are displayed in the image sensed from the land then exhibit temporal variation with seasonal periodicity. An adaptive feedback system proposed in this study reconstructs a sequence of images remotely sensed from the land surface having the physical processes with seasonal periodicity. The harmonic model is used to track seasonal variation through time, and a Gibbs random field (GRF) is used to represent the spatial dependency of digital image processes. In this study, the Normalized Difference Vegetation Index (NDVI) image was computed for one week composites of the Advanced Very High Resolution Radiometer (AVHRR) imagery over the Korean peninsula, and the adaptive reconstruction of harmonic model was then applied to the NDVI time series from 1996 to 2000 for tracking changes on the ground vegetation. The results show that the adaptive approach is potentially very effective for continuously monitoring changes on near-real time.

• Earthquakes and Structures
• /
• v.8 no.1
• /
• pp.153-184
• /
• 2015

Most damaging earthquakes come as complex sequences characterized by strong aftershocks, sometimes by foreshocks and often by multiple mainshocks. Complex earthquake sequences have enormous seismic hazard, engineering and societal implications as their impact on buildings and infrastructures may be much more severe at the end of the sequence than just after the mainshock. In this paper we examine whether historical sources can help characterizing the rare earthquake sequences of pre-instrumental times in full, including fore-, main- and aftershocks. Thanks to the its huge documentary heritage, Italy relies on one of the richest parametric earthquake catalogues worldwide. Unfortunately most current methods for assessing seismic hazard require that earthquake catalogues be declustered by removing all shocks that bear some dependency with those identified as mainshocks. We maintain that this requirement has led most modern historical seismologists to focus mainly on mainshocks rather than also on the fore- and aftershocks. To shed light onto major earthquake sequences of the past, rather than onto individual mainshocks, we investigated 10 damaging earthquake sequences ($M_w$ 4.7-7.0) that hit the L'Aquila area and central Abruzzo from the 14th to the 20th century. We find that most of the results of historical research are important for modern seismology, yet their rendering by the current parametric catalogues causes most information to be lost or not easily transferred to the potential users. For this reason we advocate a change in current strategies and the creation of a more flexible standard for storing and using all the information made available by historical seismology.

• Proceedings of the Korean Society of Crop Science Conference
• /
• 2017.06a
• /
• pp.20-20
• /
• 2017

Plant genomes include heterochromatic loci that consist of repetitive sequences and transposable elements. LTR retrotransposon is the major class of transposons in advanced plants in terms of proportion in plant genome. The elements contribute not only to genome size but also to genome stability and gene expression. A number of cases have been reported transposon insertions near genic regions affect crop traits such as fruit pigments, stress tolerance, and yields. Functional LTR retrotransposons produce extrachromosomal DNA from genomic RNA by reverse transcription that takes place within virus-like-particles (VLPs). DECREASED DNA METHYLATION 1 (DDM1) plays important roles in maintaining DNA methylation of heterochromatin affecting all sequence contexts, CG, CHG, and CHH. Previous studies showed that ddm1 mutant exhibits massive transcription of retrotransposons in Arabidopsis, but only few of them were able to create new insertions into the genome. RNA-dependent RNA POLYMERASE 6 (RDR6) is known to function in restricting accumulation of transposon RNA by processing the transcripts into 21-22 nt epigenetically activated small interfering RNA (easiRNA). We purified VLPs and sequence cDNA to identify functional LTR retrotransposons in Arabidopsis ddm1 and ddm1rdr6 plants. Over 20 LTR copia and gypsy families were detected in ddm1 and ddm1rdr6 sequencing libraries and most of them were not reported for mobility. In ddm1rdr6, short fragments of ATHILA gypsy elements were detected. It suggests easiRNAs might regulate reverse transcription steps. The highest enriched element among transposon loci was previously characterized EVADE element. It has been reported that active EVADE element is more efficiently silenced through female germline than male germline. By genetic analyses, we found ddm1 and rdr6 mutation affect maternal silencing of active EVADE elements. DDM1-GFP protein accumulated in megaspore mother cell but was not found in mature egg cell. The fusion protein was also found in early embryo and maternal DDM1-GFP allele was more dominantly expressed in the embryo. We observed localization of DDM1-GFP in Arabidopsis and DDM1-YFP in maize and found the proteins accumulated in dividing zone of root tips. Currently we are looking at cell cycle dependency of DDM1 expression using maize system. Among 10 AGO proteins in Arabidopsis, AGO9 is specifically expressed in egg cell and shoot meristematic cells. In addition, mutation of AGO9 and RDR6 caused failure in maternal silencing, implying 21-22 nt easiRNA pathway is important for retrotransposon silencing in female gametophyte or/and early embryo. On the other hand, canonical 24 nt sRNA-directed DNA methylation (RdDM) pathways did not contribute to maternal silencing as confirmed by this study. Heat-activated LTR retrotransposon, ONSEN, was not silenced by DDM1 but the silencing mechanisms require RdDM pathways in somatic cells. We will propose distinct mechanisms of LTR retrotransposons in germline and somatic stages.

• Proceedings of the Korean Society of Crop Science Conference
• /
• 2017.06a
• /
• pp.97-97
• /
• 2017

Plant genomes include heterochromatic loci that consist of repetitive sequences and transposable elements. LTR retrotransposon is the major class of transposons in advanced plants in terms of proportion in plant genome. The elements contribute not only to genome size but also to genome stability and gene expression. A number of cases have been reported transposon insertions near genic regions affect crop traits such as fruit pigments, stress tolerance, and yields. Functional LTR retrotransposons produce extrachromosomal DNA from genomic RNA by reverse transcription that takes place within virus-like-particles (VLPs). DECREASED DNA METHYLATION 1 (DDM1) plays important roles in maintaining DNA methylation of heterochromatin affecting all sequence contexts, CG, CHG, and CHH. Previous studies showed that ddm1 mutant exhibits massive transcription of retrotransposons in Arabidopsis, but only few of them were able to create new insertions into the genome. RNA-dependent RNA POLYMERASE 6 (RDR6) is known to function in restricting accumulation of transposon RNA by processing the transcripts into 21-22 nt epigenetically activated small interfering RNA (easiRNA). We purified VLPs and sequence cDNA to identify functional LTR retrotransposons in Arabidopsis ddm1 and ddm1rdr6 plants. Over 20 LTR copia and gypsy families were detected in ddm1 and ddm1rdr6 sequencing libraries and most of them were not reported for mobility. In ddm1rdr6, short fragments of ATHILA gypsy elements were detected. It suggests easiRNAs might regulate reverse transcription steps. The highest enriched element among transposon loci was previously characterized EVADE element. It has been reported that active EVADE element is more efficiently silenced through female germline than male germline. By genetic analyses, we found ddm1 and rdr6 mutation affect maternal silencing of active EVADE elements. DDM1-GFP protein accumulated in megaspore mother cell but was not found in mature egg cell. The fusion protein was also found in early embryo and maternal DDM1-GFP allele was more dominantly expressed in the embryo. We observed localization of DDM1-GFP in Arabidopsis and DDM1-YFP in maize and found the proteins accumulated in dividing zone of root tips. Currently we are looking at cell cycle dependency of DDM1 expression using maize system. Among 10 AGO proteins in Arabidopsis, AGO9 is specifically expressed in egg cell and shoot meristematic cells. In addition, mutation of AGO9 and RDR6 caused failure in maternal silencing, implying 21-22 nt easiRNA pathway is important for retrotransposon silencing in female gametophyte or/and early embryo. On the other hand, canonical 24 nt sRNA-directed DNA methylation (RdDM) pathways did not contribute to maternal silencing as confirmed by this study. Heat-activated LTR retrotransposon, ONSEN, was not silenced by DDM1 but the silencing mechanisms require RdDM pathways in somatic cells. We will propose distinct mechanisms of LTR retrotransposons in germline and somatic stages.

• Korean Journal of Microbiology
• /
• v.46 no.1
• /
• pp.80-85
• /
• 2010

Two different ${\alpha}$-amylase isozymes (AMY1 and AMY2) found in barley malt share up to 80% of amino acid sequence identity with each other, but their enzymatic properties differ remarkably. AMY1 shows the highest activity at low concentration of calcium ion, while AMY2 is highly active at high calcium concentration. Meanwhile, BASI (Barley ${\alpha}$-Amylase/Subtilisin Inhibitor) protein specifically inhibits only AMY2. In the present study, three separate regions in AMY genes (I, II, and III) were assigned on the basis of restriction enzyme sites and four kinds of chimeric amylases have been obtained by swapping a part of regions with each other. Each chimera gene was successfully over-expressed in Pichia pastoris. From the results of enzymatic characterization, both AMY211 and AMY122 showed the mixed or intermediate type of calcium-dependent activity between AMY1 and 2. Meanwhile, only AMY221 chimera could be significantly inhibited by BASI protein. As a result, it can be proposed that some amino acid residues in the region I and II, except region III, of barley ${\alpha}$-amylases play very important roles in calcium-dependency and interaction with BASI.

• Microbiology and Biotechnology Letters
• /
• v.38 no.2
• /
• pp.151-157
• /
• 2010

Barley malt produces two different $\alpha$-amylase isozymes (AMY1 and AMY2), which share up to 80% of amino acid sequence identity with each other. However, their enzymatic properties differ remarkably. In this study, five chimeric enzymes between AMY1 and 2 were constructed by staggered extension process (StEP) technique, and their enzymatic properties were characterized. According to the results, chimeric AMY-D2, D8, and E12 showed the mixed or intermediate types of calcium-dependent activity between AMY1 and 2. Meanwhile, only AMY-E10 chimera could be significantly inhibited by barley $\alpha$-amylase/subtilisin inhibitor (BASI) protein. Chimera AMY-C6 showed the same calcium-dependency as AMY1, while AMY-E10 was closely similar to AMY2. As a result, it can be proposed that some amino acid residues in the region II, III, and IV of barley $\alpha$-amylases can play very important roles in the interaction with BASI, and those in III, V, VI, and VII may partly affect on the calcium-dependent activity.

• Journal of the Korean Society of Radiology
• /
• v.18 no.1
• /
• pp.37-44
• /
• 2024

Even though MR can reveal excellent soft-tissue contrast and functional information, CT is also required for electron density information for accurate dose calculation in Radiotherapy. For the fusion of MRI and CT images in RT treatment planning workflow, patients are normally scanned on both MRI and CT imaging modalities. Recently deep-learning-based generations of CT images from MR images became possible owing to machine learning technology. This eliminated CT scanning work. This study implemented a CycleGan deep-learning-based CT image generation from MR images. Three CT generators whose learning is based on T1- , T2- , or T1-&T2-weighted MR images were created, respectively. We found that the T1-weighted MR image-based generator can generate better than other CT generators when T1-weighted MR images are input. In contrast, a T2-weighted MR image-based generator can generate better than other CT generators do when T2-weighted MR images are input. The results say that the CT generator from MR images is just outside the practical clinics and the specific weight MR image-based machine-learning generator can generate better CT images than other sequence MR image-based generators do.