• International journal of advanced smart convergence
• /
• 제5권3호
• /
• pp.8-15
• /
• 2016

In recent years, one of deep learning models called Deep Belief Network (DBN) which formed by stacking restricted Boltzman machine in a greedy fashion has beed widely used for classification and recognition. With an ability to extracting features of high-level abstraction and deal with higher dimensional data structure, this model has ouperformed outstanding result on image and speech recognition. In this research, we assess the applicability of deep learning in dna classification level. Since the training phase of DBN is costly expensive, specially if deals with DNA sequence with thousand of variables, we introduce a new encoding method, using decimal-binary vector to represent the sequence as input to the model, thereafter compare with one-hot-vector encoding in two datasets. We evaluated our proposed model with different contrastive algorithms which achieved significant improvement for the training speed with comparable classification result. This result has shown a potential of using decimal-binary vector on DBN for DNA sequence to solve other sequence problem in bioinformatics.

• 한국인터넷방송통신학회논문지
• /
• 제22권5호
• /
• pp.75-86
• /
• 2022

본 연구는 딥러닝 기반 변분 오토인코더(Variational Autoencoder)를 활용하여 수치형 학습데이터 내 클래스 불균형 문제를 해결하고, 학습데이터를 증강하여 학습모델의 성능을 향상시키고자 한다. 우리는 주어진 테이블 데이터에 대하여 인위적으로 레코드 개수를 늘리기 위해 'D-VAE'을 제안한다. 제안 기법은 최적의 데이터 증강을 지원하기 위해 우선 이산화와 특징선택을 수반한 전처리 과정을 수행한다. 이산화 과정에서 k-means 클러스터링을 적용하여 그룹화한 후, 주어진 데이터가 원-핫 인코딩(one-hot encoding) 기법으로 원-핫 벡터(one-hot vector)로 변환한다. 이후, 특징 선택 기법 중 RFECV 기법을 활용하여 예측에 도움이 되는 변수를 가려내고, 이에 대해서만 변분 오토인코더를 활용하여 새로운 학습데이터를 생성한다. 제안 기법의 성능을 검증하기 위해 4가지 유형의 실험 데이터를 활용하여 데이터 증강 비율별로 그 유효성을 입증한다.

• Journal of Life Science
• /
• 제12권2호
• /
• pp.53-56
• /
• 2002

The baculovirus expression vector system (BEVS) is one of the powerful heterologous protein expression systems using insect cells. As a result this has become a hot issue in the fleld of biotechnology. The advantage of the BEVS is that the large-scale production of heterologous proteins, which undergo posttranslational modification in the endoplasmic reticulum (ER), can be accomplished. Altrough posttranslational modification of heterologous proteins in insect cells is more similar to mammalian cells than yeast, it is not always identical. Therefore, aggregation and degradation can sometimes occur in the ER. To produce a high level of bioactive heterologous proteins using BEVS in insect cells, the prerequisite is to completely understand the posttranslational conditions that determine how newly synthesized polypeptides are folded and assembling with ER chaperones in the ER lumen. Here, we provide information on current BEVS problems and the possibility of successful heterologous protein production from mammalian cells.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2019년도 추계학술발표대회
• /
• pp.795-797
• /
• 2019

In this paper, we propose a method to generate adversarial examples for toxicity detection neural networks. Our dataset is represented by a one-hot vector and we constrain that only one character is allowed to be modified. The location to be changed is founded by the maximum area of input gradient, which represents the most affecting character the model to make decisions. Despite the fact that we have strong constraint compared to the image-based adversarial attack, we have achieved about 49% successful rate.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2011년도 춘계학술발표대회
• /
• pp.590-593
• /
• 2011

Time goes on, Ad Hoc network is hot issues. So far, there are a lot of protocols have been proposed for Ad Hoc routing protocol to support the mobility. This paper presents an enhanced DSDV(Destination-Sequenced Distance Vector) routing protocol which nominates one node to take care of a specific area. Simply Area-Aware(AA) DSDV routing protocol has one nominee to take care of some area. It has two jobs. One is to take care of its neighbour and another is to transfer the routing table to its other node as it works. It is called as Area Nominee(AN). The new scheme extends the routing table to include the nominee in the area. The general node is the same as the previous DSDV routing protocol. In the other hands, the node which is nominated has two routing protocols. One is for Regional Routing(RR) table which is the same routing table in DSDV. Another is Global Routing(GR) table which is about the area round its area which it cares nearby. GR table is the table for the designated node like the nominee. Each area has one nominee to transfer between ANs. It has only nominee's information about every area. This concept decreases the topology size and makes the information of topology more accurate.

• 천문학회보
• /
• 제36권1호
• /
• pp.35.2-35.2
• /
• 2011

We have investigated a coronal jet near the limb on 2010 June 27 by Hinode/X-Ray Telescope (XRT), EUV Imaging Spectrograph (EIS), SDO/Atmospheric Imaging Assembly (AIA), and STEREO. From EUV (AIA and EIS) and soft X-ray (XRT) images we identify the erupting jet feature in cool and hot temperatures. Using the high temporal and multi wavelength AIA images, we found that the hot jet preceded its associated cool jet and their structures are well consistent with the numerical simulation of the emerging flux-reconnection model. From the spectroscopic analysis, we found that the jet structure changes from blue shift to red one with time, which may indicate the helical structure of the jet. The STEREO observation, which enables us to observe this jet on the disk, shows that there was a dim loop associated with the jet. On the other hand, we found that the structure of its associated active region seen in STEREO is similar to that in AIA observed 5 days before. Based on this fact, we compared the jet morphology on the limb with the magnectic fields extrapolated from a HMI vector magnetogram of this active region observed on the disk. Interestingly, the comparison shows that the open and closed magnetic field configuration correspond to the jet and the dim loop, respectively, as the Shibata's jet model predicted.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2017년도 제29회 한글 및 한국어 정보처리 학술대회
• /
• pp.147-150
• /
• 2017

Deep learning의 개발에 따라 개체명 인식에도 neural network가 적용된 연구가 활발히 일어나고 있다. 영어권 개체명 인식에서는 F1 score 90%을 웃도는 성능을 내는 연구들이 나오고 있다. 하지만 한국어는 영어와 언어적 특질이 많이 달라 이를 그대로 적용시키는 데는 어려움이 있어 영어권 개체명 인식기에 비해 비교적 낮은 성능을 보인다. 본 논문에서는 "하다" 접사의 동사형이 보존된 워드 임베딩을 사용하고 한국어 개체명의 특징을 담은 one-hot 벡터를 추가하여 한국어의 특질에 보다 적합한 데이터를 deep learning 기술에 적용하였다.

• 한국어정보학회:학술대회논문집
• /
• 한국어정보학회 2017년도 제29회 한글및한국어정보처리학술대회
• /
• pp.147-150
• /
• 2017

Deep learning의 개발에 따라 개체명 인식에도 neural network가 적용된 연구가 활발히 일어나고 있다. 영어권 개체명 인식에서는 F1 score 90%을 웃도는 성능을 내는 연구들이 나오고 있다. 하지만 한국어는 영어와 언어적 특질이 많이 달라 이를 그대로 적용시키는 데는 어려움이 있어 영어권 개체명 인식기에 비해 비교적 낮은 성능을 보인다. 본 논문에서는 "하다" 접사의 동사형이 보존된 워드 임베딩을 사용하고 한국어 개체명의 특징을 담은 one-hot 벡터를 추가하여 한국어의 특질에 보다 적합한 데이터를 deep learning 기술에 적용하였다.

• Journal of information and communication convergence engineering
• /
• 제16권3호
• /
• pp.148-152
• /
• 2018

Adversarial attacks on artificial intelligence (AI) systems use adversarial examples to achieve the attack objective. Adversarial examples consist of slightly changed test data, causing AI systems to make false decisions on these examples. When used as a tool for attacking AI systems, this can lead to disastrous results. In this paper, we propose an ensemble of degraded convolutional neural network (CNN) modules, which is more robust to adversarial attacks than conventional CNNs. Each module is trained on degraded images. During testing, images are degraded using various degradation methods, and a final decision is made utilizing a one-hot encoding vector that is obtained by summing up all the output vectors of the modules. Experimental results show that the proposed ensemble network is more resilient to adversarial attacks than conventional networks, while the accuracies for normal images are similar.

• 한국전자통신학회논문지
• /
• 제15권4호
• /
• pp.725-732
• /
• 2020

본 논문은 소외계층 아동의 운동학습프로그램에서 체력 활동 중 나를 잘 따른다(0-9), 마음의 결정을 내리는데 많은 시간이 걸린다(0-9), 맥빠진(0-9) 등을 변수로 사용하여 '성별', '체육교실', 나이의 '상중하'를 분류하고 스포츠 재활치료를 통한 자아 탄력(ego-resiliency)과 자아 통제(self-control)의 변화를 관찰하여 정신 건강 변화를 알아본다. 이를 위해 취득한 데이터를 병합하고 Label encoder와 One-hot encoding을 사용하여 숫자의 크고 작음의 특성을 제거한 후 MLP, SVM, Dicesion tree, RNN, LSTM의 각각의 알고리즘을 적용하여 성능을 평가하기 위해 Train, Test 데이터를 75%, 25% 스플릿 한 뒤 Train 데이터로 알고리즘을 학습하고 Test 데이터로 알고리즘의 정확성을 측정한다. 측정 결과 성별에서는 LSTM, 체육 교실은 MLP와 LSTM, 나이는 SVM이 가장 우수한 결과를 보임을 확인하였다.