• ETRI Journal
• /
• v.43 no.5
• /
• pp.775-786
• /
• 2021

To translate in real time, a simultaneous translation system should determine when to stop reading source tokens and generate target tokens corresponding to a partial source sentence read up to that point. However, conventional attention-based neural machine translation (NMT) models cannot produce translations with adequate latency in online scenarios because they wait until a source sentence is completed to compute alignment between the source and target tokens. To address this issue, we propose a reinforced learning (RL)-based attention mechanism, the reinforced attention mechanism, which allows a neural translation model to jointly train the stopping criterion and a partial translation model. The proposed attention mechanism comprises two modules, one to ensure translation quality and the other to address latency. Different from previous RL-based simultaneous translation systems, which learn the stopping criterion from a fixed NMT model, the modules can be trained jointly with a novel reward function. In our experiments, the proposed model has better translation quality and comparable latency compared to previous models.

• Annual Conference on Human and Language Technology
• /
• 2013.10a
• /
• pp.70-73
• /
• 2013

자연어처리에서는 많은 모듈들이 파이프라인 방식으로 연결되어 사용되나, 이 경우 앞 단계의 오류가 뒷 단계에 누적되는 문제와 앞 단계에서 뒷 단계의 정보를 사용하지 못한다는 단점이 있다. 본 논문에서는 파이프라인 방식의 문제를 해결하기 위해 사용되는 일반적인 결합 학습 방법을 확장하여, 두 작업이 동시에 태깅된 학습 데이터뿐만 아니라 한 작업만 태깅된 학습데이터도 동시에 학습에 사용할 수 있는 결합 학습 모델을 Latent Structural SVM을 확장하여 제안한다. 실험 결과, 기존의 한국어 띄어쓰기와 품사 태깅 결합 모델의 품사 태깅 성능이 96.99%였으나, 본 논문에서 제안하는 결합 학습 모델을 이용하여 대용량의 한국어 띄어쓰기 학습데이터를 추가로 학습한 결과 품사 태깅 성능이 97.20%까지 향상 되었다.

• International Journal of Advanced Culture Technology
• /
• v.10 no.4
• /
• pp.499-510
• /
• 2022

This study examines the reporting factors of crime against business in Korea and proposes a corresponding predictive model using machine learning. While many previous studies focused on the individual factors of theft victims, there is a lack of evidence on the reporting factors of crime against a business that serves the public good as opposed to those that protect private property. Therefore, we proposed a crime prevention model for the willingness factor of theft reporting in businesses. This study used data collected through the 2015 Commercial Crime Damage Survey conducted by the Korea Institute for Criminal Policy. It analyzed data from 834 businesses that had experienced theft during a 2016 crime investigation. The data showed a problem with unbalanced classes. To solve this problem, we jointly applied the Synthetic Minority Over Sampling Technique and the Tomek link techniques to the training data. Two prediction models were implemented. One was a statistical model using logistic regression and elastic net. The other involved a support vector machine model, tree-based machine learning models (e.g., random forest, extreme gradient boosting), and a stacking model. As a result, the features of theft price, invasion, and remedy, which are known to have significant effects on reporting theft offences, can be predicted as determinants of such offences in companies. Finally, we verified and compared the proposed predictive models using several popular metrics. Based on our evaluation of the importance of the features used in each model, we suggest a more accurate criterion for predicting var.

• Journal of Korean Elementary Science Education
• /
• v.25 no.1
• /
• pp.8-14
• /
• 2006

This study examined how pre-service elementary teachers' perceptions regarding the learning environment (learning focus/ability-meritocracy/cooperative climate) and achievement goals (mastery/performance-approach/performance-avoidance) in science education courses jointly contributed to their science teaching efficacy beliefs (personal science teaching efficacy belief/science teaching outcome expectancy). A path analysis supported a causal model in which the perception of the learning focus influenced the mastery goal, which in turn influenced the personal science teaching efficacy belief and science teaching outcome expectancy. The perception of learning focus also had a direct effect on science teaching outcome expectancy. The perception of ability-meritocracy influenced personal science teaching efficacy belief via the performance-approach (positively) or, conversely, the performance-avoidance goal (negatively). No link .was deduced from the perception of cooperative climate. The educational implications of these findings were also discussed.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.113-117
• /
• 2020

Images were taken under various weather such as rain, haze, snow often show low visibility, which can dramatically decrease accuracy of some tasks in computer vision: object detection, segmentation. Besides, previous work to enhance image usually downsample the image to receive consistency features but have not yet good upsample algorithm to recover original size. So, in this research, we jointly implement removal streak in heavy rain image and super resolution using a deep network. We put forth a 2-stage network: a multi-model network followed by a refinement network. The first stage using rain formula in the single image and two operation layers (addition, multiplication) removes rain streak and noise to get clean image in low resolution. The second stage uses refinement network to recover damaged background information as well as upsample, and receive high resolution image. Our method improves visual quality image, gains accuracy in human action recognition task in datasets. Extensive experiments show that our network outperforms the state of the art (SoTA) methods.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.3
• /
• pp.647-654
• /
• 2023

Deep Learning is a useful method for classifying and recognizing complex data such as images and text, and the accuracy of the deep learning method is the basis for making artificial intelligence-based services on the Internet useful. However, the vast amount of user da vita used for training in deep learning has led to privacy violation problems, and it is worried that companies that have collected personal and sensitive data of users, such as photographs and voices, own the data indefinitely. Users cannot delete their data and cannot limit the purpose of use. For example, data owners such as medical institutions that want to apply deep learning technology to patients' medical records cannot share patient data because of privacy and confidentiality issues, making it difficult to benefit from deep learning technology. In this paper, we have designed a privacy preservation technique-applied deep learning technique that allows multiple workers to use a neural network model jointly, without sharing input datasets, in multi-party system. We proposed a method that can selectively share small subsets using an optimization algorithm based on modified stochastic gradient descent, confirming that it could facilitate training with increased learning accuracy while protecting private information.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.1
• /
• pp.392-412
• /
• 2018

Due to the semantic gap problem across different modalities, automatically retrieval from multimedia information still faces a main challenge. It is desirable to provide an effective joint model to bridge the gap and organize the relationships between them. In this work, we develop a deep image annotation and classification by fusing multi-modal semantic topics (DAC_mmst) model, which has the capacity for finding visual and non-visual topics by jointly modeling the image and loosely related text for deep image annotation while simultaneously learning and predicting the class label. More specifically, DAC_mmst depends on a non-parametric Bayesian model for estimating the best number of visual topics that can perfectly explain the image. To evaluate the effectiveness of our proposed algorithm, we collect a real-world dataset to conduct various experiments. The experimental results show our proposed DAC_mmst performs favorably in perplexity, image annotation and classification accuracy, comparing to several state-of-the-art methods.

• Journal of Soil and Groundwater Environment
• /
• v.21 no.4
• /
• pp.30-41
• /
• 2016

The amount of groundwater related data is drastically increasing domestically from various sources since 2000. To justify the more expansive continuation of the data acquisition and to derive valuable implications from the data, continued employments of sophisticated and state-of-the-arts statistical tools in the analyses and predictions are important issue. In the present study, we employed a well established machine learning technique of Gaussian Process Regression (GPR) model in the trend analyses of groundwater level for the long-term change. The major benefit of GPR model is that the model provide not only the future predictions but also the associated uncertainty. In the study, the long-term predictions of groundwater level from the stations of National Groundwater Monitoring Network located within Han River Basin were exemplified as prediction cases based on the GPR model. In addition, a few types of groundwater change patterns were delineated (i.e., increasing, decreasing, and no trend) on the basis of the statistics acquired from GPR analyses. From the study, it was found that the majority of the monitoring stations has decreasing trend while small portion shows increasing or no trend. To further analyze the causes of the trend, the corresponding precipitation data were jointly analyzed by the same method (i.e., GPR). Based on the analyses, the major cause of decreasing trend of groundwater level is attributed to reduction of precipitation rate whereas a few of the stations show weak relationship between the pattern of groundwater level changes and precipitation.

• Journal of the Society of Disaster Information
• /
• v.19 no.3
• /
• pp.665-672
• /
• 2023

Purpose: Underground utility tunnel is facility that is jointly house infrastructure such as electricity, water and gas in city, causing condensation problems due to lack of airflow. This paper aims to prevent electricity leakage fires caused by condensation by detecting whether the control panel door in the underground utility tunnel is open using a deep learning model. Method: YOLO, a deep learning object recognition model, is trained to recognize the opening and closing of the control panel door using video data taken by a robot patrolling the underground utility tunnel. To improve the recognition rate, image augmentation is used. Result: Among the image enhancement techniques, we compared the performance of the YOLO model trained using mosaic with that of the YOLO model without mosaic, and found that the mosaic technique performed better. The mAP for all classes were 0.994, which is high evaluation result. Conclusion: It was able to detect the control panel even when there were lights off or other objects in the underground cavity. This allows you to effectively manage the underground utility tunnel and prevent disasters.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.2
• /
• pp.15-23
• /
• 2017

The person Re-identification is the most challenging part of computer vision due to the significant changes in human pose and background clutter with occlusions. The picture from non-overlapping cameras enhance the difficulty to distinguish some person from the other. To reach a better performance match, most methods use feature selection and distance metrics separately to get discriminative representations and proper distance to describe the similarity between person and kind of ignoring some significant features. This situation has encouraged us to consider a novel method to deal with this problem. In this paper, we proposed an enhanced recurrent neural network with three-tier hierarchical network for person re-identification. Specifically, the proposed recurrent neural network (RNN) model contain an iterative expectation maximum (EM) algorithm and three-tier Hierarchical network to jointly learn both the discriminative features and metrics distance. The iterative EM algorithm can fully use of the feature extraction ability of convolutional neural network (CNN) which is in series before the RNN. By unsupervised learning, the EM framework can change the labels of the patches and train larger datasets. Through the three-tier hierarchical network, the convolutional neural network, recurrent network and pooling layer can jointly be a feature extractor to better train the network. The experimental result shows that comparing with other researchers' approaches in this field, this method also can get a competitive accuracy. The influence of different component of this method will be analyzed and evaluated in the future research.