學術出版;;Publications

Permanent URI for this communityhdl:11536/1

收集所有交通大學研究人員與學生個人或團體所發表之各項學術著作，涵蓋期刊論文、教師專書、學位論文等項。;;Collections of Publications of NCTU Researchers and Students Like Journal Papers and Thesis.

Browse

🏛️國立陽明交通大學機構典藏

Search Results

Now showing 1 - 10 of 23

A two-step deep learning approach to data classification and modeling and a demonstration on subject type relationship analysis in the Web of Science
(1970-01-01)
Phoa, Frederick Kin Hing
;
Lai, Hsin-Yi
;
Chang, Livia Lin-Hsuan
;
Honda, Keisuke
;
統計學研究所
It is common sense that some subjects have strong relationships while others are perhaps almost mutually independent, but a quantitative and systematic approach to describe such sense is a deficiency. A technique called pointwise mutual information (PMI) from information science helps to fulfill the request, but the calculation through a large-scale database is computationally infeasible if one requires an instantaneous value. This work provides a two-step remedy via deep learning for estimating and predicting relationships among two subject types that are found in the large-scale citation database called the Web of Science. The resulting model successfully replicates existing PMI values among subject types, and it can be used for predicting PMI values of two subject types if one or both subject types does not exist in the database.
A Survey on Deep Learning-Based Vehicular Communication Applications
(1970-01-01)
Lin, Chia-Hung
;
Lin, Yu-Chien
;
Wu, Yen-Jung
;
Chung, Wei-Ho
;
Lee, Ta-Sung
Besides the use of information transmission, vehicular communications also perform an essential role in intelligent transportation systems (ITS) for exchanging critical driving information among end users, vehicles, and infrastructures. Moreover, to enhance the understanding of the local environment, increasingly more data are collected by sensors, inducing an extensive use of deep learning (DL)-based algorithms in ITS. To further promote the development of DL-based algorithms in ITS, in this paper, we present a concise introduction of DL technologies. Then, we conduct an in-depth investigation on two popular DL-based applications used in ITS, traffic flow forecasting and trajectory prediction, focusing onwhenandhowthe authors employ different DL models and training schemes in these tasks. Finally, we raise two existing problems while employing DL-based algorithms in practical ITS and further discuss certain recent advances in DL-based research to tackle these challenges. To encourage more researchers to focus on the development of DL-based algorithms in ITS for a better world, we hope this paper can be treated as an informational material for prospective researchers, which contains the essential background knowledge of DL-based ITS applications; we also hope this paper will encourage experienced researchers to counter the open challenges and achieve a technical breakthrough to ITS.
Deep learning and control algorithms of direct perception for autonomous driving
(1970-01-01)
Lee, Der-Hau
;
Chen, Kuan-Lin
;
Liou, Kuan-Han
;
Liu, Chang-Lun
;
Liu, Jinn-Liang
We propose an end-to-end machine learning model that integrates multi-task (MT) learning, convolutional neural networks (CNNs), and control algorithms to achieve efficient inference and stable driving for self-driving cars. The CNN-MT model can simultaneously perform regression and classification tasks for estimating perception indicators and driving decisions, respectively, based on the direct perception paradigm of autonomous driving. The model can also be used to evaluate the inference efficiency and driving stability of different CNNs on the metrics of CNN's size, complexity, accuracy, processing speed, and collision number, respectively, in a dynamic traffic. We also propose new algorithms for controllers to drive a car using the indicators and its short-range sensory data to avoid collisions in real-time testing. We collect a set of images from a camera of The Open Racing Car Simulator in various driving scenarios, train the model using this dataset, test it in unseen traffics, and find that it outperforms earlier models in highway traffic. The stability of end-to-end learning and self driving depends crucially on the dynamic interplay between CNN and control algorithms. The source code and data of this work are available on our website, which can be used as a simulation platform to evaluate different learning models on equal footing and quantify collisions precisely for further studies on autonomous driving.
Convolutional neural networks for classification of music-listening EEG: comparing 1D convolutional kernels with 2D kernels and cerebral laterality of musical influence
(2020-07-01)
Cheah, Kit Hwa
;
Nisar, Humaira
;
Yap, Vooi Voon
;
Lee, Chen-Yi
;
電子工程學系及電子研究所
This paper highlights the ability of convolutional neural networks (CNNs) at classifying EEG data listening to different kinds of music without the requirement for handcrafted features. Deep learning architectures presented in this paper include CNN of different depths and different convolutional kernels. Support vector machine (SVM) taking in EEG features describing the frequency spectrum, signal regularity, and cross-channel correlation has been applied for performance comparison with CNN. The best performing CNN model presented in this paper achieves the tenfold cross-validation (CV) binary classification average accuracy of 98.94% (validation) and 97.46% (test), and the tenfold CV three-class classification accuracy of 97.68% (validation) and 95.71% (test). In comparison, the SVM classifier achieves tenfold CV binary classification accuracy of 80.23% (validation). The CNN model presented is able to not only differentiate EEG of subjects listening to music from that of subjects without auditory input, but it is also capable of accurately differentiating the EEG of subjects listening to different music. In the context of designing neural computing models for EEG analysis, this paper shows that decomposing two-dimensional spatiotemporal convolutional kernels into separate one-dimensional spatial and one-dimensional temporal kernels significantly reduces the number of trainable parameters (size) of the model while retaining the classification performance. This finding is useful, especially in designing CNN for memory-critical embedded systems for EEG processing. In neurological aspect, auditory stimulus is found to have altered the EEG pattern of the frontal lobe and the left cerebral hemisphere more than the other brain regions.
A dimension-reduction based multilayer perception method for supporting the medical decision making
(2020-03-01)
Lee, Shin-Jye
;
Tseng, Ching-Hsun
;
Lin, G. T-R
;
Yang, Yun
;
Yang, Po
Due to the rapid development of Medical IoT recently, how to effectively apply these huge amounts of IoT data to enhance the reliability of the clinical decision making has become an increasing issue in the medical field. These data usually comprise high-complicated features with tremendous volume, and it implies that the simple inference models may less powerful to be practiced. In deep learning, multilayer perceptron (MLP) is a kind of feed-forward artificial neural network, and it is one of the high-performance methods about stochastic scheme, fitness approximation, and regression analysis. To process these high uncertain data, the proposed work based on MLP structure in particular integrates the boosting scheme and dimension-reduction process. In this proposed work, the advanced ReLU-based activation function is used. Also, the weight initialization is applied to improve the stable prediction and convergence. After the improved dimension-reduction process is introduced, the proposed method can effectively learn the hidden information from the reformative data and the precise labels also can be recognized by stacking a small amount of neural network layers with paying few extra cost. The proposed work shows a possible path of embedding dimension reduction in deep learning structure with minor price. In addition to the prediction issue, the proposed method can also be applied to assess risk and forecast trend among different information systems. (C) 2019 Elsevier B.V. All rights reserved.
Intelligent compilation of patent summaries using machine learning and natural language processing techniques
(2020-01-01)
Trappey, Amy J. C.
;
Trappey, Charles V.
;
Wu, Jheng-Long
;
Wang, Jack W. C.
;
管理科學系
Patents are a type of intellectual property with ownership and monopolistic rights that are publicly accessible published documents, often with illustrations, registered by governments and international organizations. The registration allows people familiar with the domain to understand how to re-create the new and useful invention but restricts the manufacturing unless the owner licenses or enters into a legal agreement to sell ownership of the patent. Patents reward the costly research and development efforts of inventors while spreading new knowledge and accelerating innovation. This research uses artificial intelligence natural language processing, deep learning techniques and machine learning algorithms to extract the essential knowledge of patent documents within a given domain as a means to evaluate their worth and technical advantage. Manual patent abstraction is a time consuming, labor intensive, and subjective process which becomes cost and outcome ineffective as the size of the patent knowledge domain increases. This research develops an intelligent patent summarization methodology using artificial intelligence machine learning approaches to allow patent domains of extremely large sizes to be effectively and objectively summarized, especially for cases where the cost and time requirements of manual summarization is infeasible. The system learns to automatically summarize patent documents with natural language texts for any given technical domain. The machine learning solution identifies technical key terminologies (words, phrases, and sentences) in the context of the semantic relationships among training patents and corresponding summaries as the core of the summarization system. To ensure the high performance of the proposed methodology, ROUGE metrics are used to evaluate precision, recall, accuracy, and consistency of knowledge generated by the summarization system. The Smart machinery technologies domain, under the sub-domains of control intelligence, sensor intelligence and intelligent decision-making provide the case studies for the patent summarization system training. The cases use 1708 training pairs of patents and summaries while testing uses 30 randomly selected patents. The case implementation and verification have shown the summary reports achieve 90% and 84% average precision and recall ratios respectively.
DEEP LEARNING-BASED OBSTACLE DETECTION AND DEPTH ESTIMATION
(2019-01-01)
Hsieh, Yi-Yu
;
Lin, Wei-Yu
;
Li, Dong-Lin
;
Chuang, Jen-Hui
;
資訊工程學系
This paper proposed a modified YOLOv3 which has an extra object depth prediction module for obstacle detection and avoidance. We use a pre-processed KITTI dataset to train the proposed, unified model for (i) object detection and (ii) depth prediction and use the AirSim flight simulator to generate synthetic aerial images to verify that our model can be applied in different data domains. Experimental results show that the proposed model compares favorably with other depth map prediction methods in terms of accuracy in the prediction of object depth for pre-processed KITTI dataset, while the unified approach can actually improve both (i) and (ii) at the same time.
USING C3D TO DETECT REAR OVERTAKING BEHAVIOR
(2019-01-01)
Tseng, Ching-Kai
;
Liao, Chien-Chih
;
Shen, Po-Chun
;
Guo, Jiun-In
;
電子工程學系及電子研究所
Avoiding traffic accidents is critical since the death of traffic accidents is the eighth among the top ten leading causes of death in 2018. This paper proposes a light-weight convolutional 3D (C3D) network with five 3D convolution layers and two fully-connected layers to predict overtaking behavior. This network utilizes the last layer of convolution layer to learn the overtaking object location in the final frame. Based on NVIDIA Jetson TX2, the proposed C3D network achieves 91.46% accuracy to detect overtaking behavior on rainy days. To generate this excellent deep learning model, we use an efficient labeling tool, called ezLabel, which is a free SaaS for academia group with 96,000 opened image data samples for deep learning. ezLabel owns outstanding route prediction and fitting functions, which speeds up with the factor of ten compared to traditional tools. Users only label the object in its first frame and in its final frame, and then ezLabel labels the object in all frames in between and fits the bounding box to the object. The ezLabel can be used to label objects captured with any moving or static cameras efficiently.
A Deep Learning-Based Strategy to the Energy Management-Advice for Time-of-Use Rate of Household Electricity Consumption
(2020-01-01)
Wu, Lu-Xian
;
Lee, Shin-Jye
;
科技管理研究所
;
Institute of Management of Technology
With the high industrialization at a rapid pace, the demand for energy increases exponentially, but it is difficult to meet the balance of demand and supply. Therefore, how to effectively meet the balance of demand and supply intelligently has become a popular issue in this century. In recent years, Taiwan Power Company (Taipower), the biggest electric company in Taiwan, is committed to the construction of Advanced Metering Infrastructure (AMI), which provides communication channels and enables demand-side users to participate in load dispatch. On the other hand, the construction of AMI is expected to generate a tremendous number of valuable data on electricity consumption, but it is not easy to convert these data into effective information by the conventional quantitative methods. In as much as the rapid progression of AI technology in the industrial field, the application of AI technology in the technology management has become an increasing issue as an interdisciplinary study. To address this task, this work applies the recurrent neural network based on deep learning to predict low-voltage usage shortly by the electricity information of low-voltage user and meteorological data. After many vicissitudes, the electricity consumption per hour can be predicted and a sound energy arrangement can be therefore planned. Through introducing the proposed model, Taipower Company will have an effective capability that schedules power, reduces unnecessary backup power, and provides time-consuming electricity prices for industrial enterprises accurately among high usage of Taiwan industries.
Mobility Predictions for IoT Devices Using Gated Recurrent Unit Network
(2020-01-01)
Belay, Abebe
;
Lin, Hsin-Piao
;
Wang, Li-Chun
;
電機工程學系
;
Department of Electrical and Computer Engineering
Wireless and mobile technologies, as well as their users, are growing rapidly as the Internet of Things (IoT) products, such as sensor-network technologies, mobile devices, and supporting applications, become widely dispersed. Owing to the dynamic changes in the wireless networks and the exponential growth of the IoT products, which make it difficult to locate large quantities of users and devices, providing accurate tracking and trajectory predictions in open and highly condensed wireless networks is extremely difficult. An adaptive and scalable system is required to offer accurate location-based services (LBSs) for the success of IoT. To enhance the attainment of IoT, we propose a hybrid of principal component analysis (PCA) and gated recurrent unit (GRU) algorithms for mobility predictions in a wireless urban area. During the system development processes, we first collect an LTE signal from three unmanned aerial vehicle base stations (UAV-BSs), the Wi-Fi signal strength from each reachable Wi-Fi access points (APs), and channel information from the Wi-Fi signal media. We then apply PCA to reduce the number of Wi-Fi features and to decrease signal noise. Next, we train the GRU algorithm to develop models that can predict the mobility of IoT device users. Finally, we evaluate the tracking and trajectory models. To evaluate the proposed techniques, we compare the common parameters of the GRU with those of other deep learning types. The proposed technique provides plausible and state-of-the-art results for mobility predictions of IoT devices in a wireless environment.

學術出版;;Publications

Browse

Filters

Settings

Sort By

Results per page

🏛️國立陽明交通大學機構典藏

Search Results