• 한국지능정보시스템학회:학술대회논문집
• /
• 한국지능정보시스템학회 2000년도 추계정기학술대회:지능형기술과 CRM
• /
• pp.427-436
• /
• 2000

The prediction of stock price index is a very difficult problem because of the complexity of the stock market data it data. It has been studied by a number of researchers since they strong1y affect other economic and financial parameters. The movement of stock price index has a series of change points due to the strategies of institutional investors. This study presents a two-stage forecasting model of stock price index using change-point detection and artificial neural networks. The basic concept of this proposed model is to obtain Intervals divided by change points, to identify them as change-point groups, and to use them in stock price index forecasting. First, the proposed model tries to detect successive change points in stock price index. Then, the model forecasts the change-point group with the backpropagation neural network (BPN). Fina1ly, the model forecasts the output with BPN. This study then examines the predictability of the integrated neural network model for stock price index forecasting using change-point detection.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권12호
• /
• pp.4439-4455
• /
• 2021

Compared with two-stage object detection algorithms, one-stage algorithms provide a better trade-off between real-time performance and accuracy. However, these methods treat the intermediate features equally, which lacks the flexibility to emphasize meaningful information for classification and location. Besides, they ignore the interaction of contextual information from different scales, which is important for medium and small objects detection. To tackle these problems, we propose an image pyramid network based on dual attention mechanism (DAIPNet), which builds an image pyramid to enrich the spatial information while emphasizing multi-scale informative features based on dual attention mechanisms for one-stage object detection. Our framework utilizes a pre-trained backbone as standard detection network, where the designed image pyramid network (IPN) is used as auxiliary network to provide complementary information. Here, the dual attention mechanism is composed of the adaptive feature fusion module (AFFM) and the progressive attention fusion module (PAFM). AFFM is designed to automatically pay attention to the feature maps with different importance from the backbone and auxiliary network, while PAFM is utilized to adaptively learn the channel attentive information in the context transfer process. Furthermore, in the IPN, we build an image pyramid to extract scale-wise features from downsampled images of different scales, where the features are further fused at different states to enrich scale-wise information and learn more comprehensive feature representations. Experimental results are shown on MS COCO dataset. Our proposed detector with a 300 × 300 input achieves superior performance of 32.6% mAP on the MS COCO test-dev compared with state-of-the-art methods.

• International Journal of Contents
• /
• 제16권2호
• /
• pp.8-18
• /
• 2020

Data is critical in deep learning but the scarcity of data often occurs in research, especially in the preparation of the paired training data. In this paper, document image binarization with unpaired data is studied by introducing adversarial learning, excluding the need for supervised or labeled datasets. However, the simple extension of the previous unpaired training to binarization inevitably leads to poor performance compared to paired data training. Thus, a new deep learning approach is proposed by introducing a multi-diversity of higher quality generated images. In this paper, a two-stage model is proposed that comprises the generative adversarial network (GAN) followed by the U-net network. In the first stage, the GAN uses the unpaired image data to create paired image data. With the second stage, the generated paired image data are passed through the U-net network for binarization. Thus, the trained U-net becomes the binarization model during the testing. The proposed model has been evaluated over the publicly available DIBCO dataset and it outperforms other techniques on unpaired training data. The paper shows the potential of using unpaired data for binarization, for the first time in the literature, which can be further improved to replace paired data training for binarization in the future.

• 대한임베디드공학회논문지
• /
• 제15권3호
• /
• pp.119-127
• /
• 2020

The detection and recognition of abnormal driving becomes crucial for achieving safety in Intelligent Transportation Systems (ITS). This paper presents a feature extraction method based on spectral data to train a neural network model for driving behavior recognition. The proposed method uses a two stage signal processing approach to derive time-saving and efficient feature vectors. For the first stage, the feature vector set is obtained by calculating variances from each frequency bin containing the power spectrum data. The feature set is further reduced in the second stage where an intersection method is used to select more significant features that are finally applied for training a neural network model. A stream of live signals are fed to the trained model which recognizes the abnormal driving behaviors. The driving behaviors considered in this study are weaving, sudden braking and normal driving. The effectiveness of the proposed method is demonstrated by comparing with existing methods, which are Particle Swarm Optimization (PSO) and Convolution Neural Network (CNN). The experiments show that the proposed approach achieves satisfactory results with less computational complexity.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권7호
• /
• pp.2547-2567
• /
• 2021

It has been widely acknowledged that occlusion impairments adversely distress many face recognition algorithms' performance. Therefore, it is crucial to solving the problem of face image occlusion in face recognition. To solve the image occlusion problem in face recognition, this paper aims to automatically de-occlude the human face majority or discriminative regions to improve face recognition performance. To achieve this, we decompose the generative process into two key stages and employ a separate generative adversarial network (GAN)-based network in both stages. The first stage generates an initial coarse face image without an occlusion mask. The second stage refines the result from the first stage by forcing it closer to real face images or ground truth. To increase the performance and minimize the artifacts in the generated result, a new refine loss (e.g., reconstruction loss, perceptual loss, and adversarial loss) is used to determine all differences between the generated de-occluded face image and ground truth. Furthermore, we build occluded face images and corresponding occlusion-free face images dataset. We trained our model on this new dataset and later tested it on real-world face images. The experiment results (qualitative and quantitative) and the comparative study confirm the robustness and effectiveness of the proposed work in removing challenging occlusion masks with various structures, sizes, shapes, types, and positions.

• 한국수자원학회논문집
• /
• 제43권2호
• /
• pp.131-138
• /
• 2010

다양화되는 물 수요와 기상 이변 등의 영향으로 극심해지는 가뭄에 대비하여 대체 수자원의 확보는 수자원 연구의 매우 중요한 부분이 되었다. 다양한 대체 수자원 중 하수처리장의 방류수는 양호한 수질과 비교적 예측이 가능한 방류량으로 인해 농업용수나 공업용수 혹은 공공용수를 대체할 안정적인 수원으로 관심의 대상이 되고 있다. 본 연구에서는 하수처리수 재이용을 위해 미래의 불확실한 용수 수요량을 고려한 최소의 공사비를 최적화하는 방법을 이진변수를 가지는 2단계 추계학적 선형계획법을 이용하여 제시하였다. 현재 설계하는 하수처리수 재이용 모형은 미래의 용수 수요량까지 고려하여 설계하여야 한다는 점을 고려하여, 미래에 용수수요가 증가할 경우, 기존의 관에 평행한 다른 관을 추가로 건설할 수 있다고 가정하여 2단계에 걸쳐 공사가 가능한 모형을 구축하였다. 그 결과 미래의 물 사용량까지를 모두 고려하여 현재 큰 직경의 관로를 건설하는 경우와 작은 직경의 관로를 두 번에 걸쳐 건설하는 대안 사이의 비용차이를 고려한 모형이 제안되었으며, 가상의 네트워크에 적용되어 그 적용성을 입증하였다. 제안된 모형은 하수 처리수 재이용 네트워크 계획 시 경제적인 관로 설계를 위한 기본 자료로 활용될 수 있으며, 장기적인 물 공급 계획을 수립할 시 여러 가지 설계 대안들에 대한 비교를 위해도 사용이 가능하다.

• International Journal of Internet, Broadcasting and Communication
• /
• 제11권4호
• /
• pp.76-85
• /
• 2019

We introduce a vision-based object detection method for real-time video surveillance system in low-end edge computing environments. Recently, the accuracy of object detection has been improved due to the performance of approaches based on deep learning algorithm such as Region Convolutional Neural Network(R-CNN) which has two stage for inferencing. On the other hand, one stage detection algorithms such as single-shot detection (SSD) and you only look once (YOLO) have been developed at the expense of some accuracy and can be used for real-time systems. However, high-performance hardware such as General-Purpose computing on Graphics Processing Unit(GPGPU) is required to still achieve excellent object detection performance and speed. To address hardware requirement that is burdensome to low-end edge computing environments, We propose sub-frame analysis method for the object detection. In specific, We divide a whole image frame into smaller ones then inference them on Convolutional Neural Network (CNN) based image detection network, which is much faster than conventional network designed forfull frame image. We reduced its computationalrequirementsignificantly without losing throughput and object detection accuracy with the proposed method.

• 대한전자공학회논문지
• /
• 제25권8호
• /
• pp.972-982
• /
• 1988

In multiprocessor systems, there are Omega network and M network among various MIN's which interconnect the processor and memory modules. Both one-path Omega network and two-path M network are composed of Log2N stages. In this paper, Augmented M network (AMN) with 2**k+1 paths and Augmented Omega network (AON) with 2**k paths are proposed. The proposed networks can be acomplished by adding K stage(s) to M network and Omega network. Using destination tag, routing algorithm for AMN and AON becomes simple and multiple faults are tolerant. By evaluating RST(request service time) performance of AMN and AON with (Log2N)+K stages, we demonstrated the fact that MMIN (AMN) with 2**k+1 paths performs better than MMIN(AON) with 2**k+1. paths.

• Water Engineering Research
• /
• 제1권1호
• /
• pp.63-74
• /
• 2000

This paper presents a two- phase search scheme for optimal pipe expansion of expansion of existing water distribution systems. In pipe network problems, link flows affect the total cost of the system because the link flows are not uniquely determined for various pipe diameters. The two-phase search scheme based on stochastic optimization scheme is suggested to determine the optimal link flows which make the optimal design of existing pipe network. A sample pipe network is employed to test the proposed method. Once the best tree network is obtained, the link flows are perturbed to find a near global optimum over the whole feasible region. It should be noted that in the perturbation stage the loop flows obtained form the sample existing network are employed as the initial loop flows of the proposed method. It has been also found that the relationship of cost-hydraulic gradient for pipe expansion of existing network affects the total cost of the sample network. The results show that the proposed method can yield a lower cost design than the conventional design method and that the proposed method can be efficiently used to design the pipe expansion of existing water distribution systems.

• 산업경영시스템학회지
• /
• 제44권3호
• /
• pp.248-261
• /
• 2021

Scientific and technological performances (e.g., patents and publications) made through R&D play a pivotal role for national economic growth. National governments encourage academia-industry cooperation and thereby pursue continuous development of science technology and innovation. Increasing R&D-related investments and manpower are crucial for national industrial development, but evidence of poor performance in business performance, efficiency, and effectiveness has recently been found in Korea. This study evaluates performance efficiency of the 6T sector (Information Technology, Bio Technology, Nano Technology, Space Technology, Environment Technology, Culture Technology), which is considered a high-potential promising industry for the next generation growth and currently occupies two thirds of the national R&D projects. The study measures the relative efficiency of R&D in a comparative perspective by employing the Data Envelopment Analysis (DEA) method. The result reveals overall low efficiency in basic R&D (0.2112), applied R&D (0.2083), development R&D (0.2638), and others (0.0641), confirming that economic performance and efficiency were relatively poor compared to production efficiency. Efficient R&D needs policy makers to create strategies that can increase overall efficiency by improving productivity performance and quality while increasing economic performance.