检索结果-内蒙古大学图书馆

Triangular architecture for rare language translation

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Ren, Shuo Chen, Wenhu Liu, Shujie Li, Mu Zhou, Ming Ma, Shuai SKLSDE Lab Beihang University Beijing China Beijing Advanced Innovation Center for Big Data and Brain Computing Beijing China University of California Santa BarbaraCA United States Microsoft Research in Asia Beijing China

Neural Machine Translation (NMT) performs poor on the low-resource language pair (X;Z), especially when Z is a rare language. By introducing another rich language Y , we propose a novel triangular training architecture (TA-NMT) to leverage bilingual data (Y;Z) (may be small) and (X;Y ) (can be rich) to improve the translation performance of lowresource pairs. In this triangular architecture, Z is taken as the intermediate latent variable, and translation models of Z are jointly optimized with a unified bidirectional EM algorithm under the goal of maximizing the translation likelihood of (X;Y ). Empirical results demonstrate that our method significantly improves the translation quality of rare languages on MultiUN and IWSLT2012 datasets, and achieves even better performance combining back-translation methods. Copyright © 2018, The Authors. All rights reserved.

关键词： Neural machine translation

An Improved Convolutional Neural Network Algorithm for Multi-Label Classification

学校读者我要写书评

暂无评论

An Improved Convolutional Neural Network Algorithm for Multi...

International Conference on Audio, Language and Image Processing, ICALIP

作者： Xinsheng Wang Lijun Sun Zhihua Wei Department of Computer Science and Technology Tongji University Shanghai China Research Center of Big Data and Network Security Tongji University Shanghai China Key Laboratory of Embedded System and Service Computing Tongji University Shanghai China

Recent years conventional neural network(CNN) has been applied to different natural language processing(NLP) tasks such as sentence classification, sentence modeling, etc. Some researchers use CNN to do multi-label classification but their work mainly focus on image rather than text. In this paper, we propose an improved CNN via hierarchical dirichlet process(HDP) model to deal with the multi-label classification problem in NLP. We first apply an HDP model to discard some words which are less important semantically. Then we use word embedding methods to transform words to vectors. Finally, we train CNN based on word vectors. Experimental results demonstrate that our method is superior to most traditional multi-label classification methods and TextCNN in terms of performance.

关键词： Feature extraction Training Convolutional neural networks Natural language processing Vocabulary Transforms

Educational mobile game for learning English words

学校读者我要写书评

暂无评论

IOP Conference Series: Materials Science and Engineering 2020年第1期979卷

作者： B K Ng N M Suaib A J Sihes A Ali Z A Shah School of Computing Faculty of Engineering Universiti Teknologi Malaysia 81310 UTM Skudai Johor Malaysia. UTM Big Data Center Ibnu Sina Institute of Scientific and Industrial Research Universiti Teknologi Malaysia 81310 UTM Skudai Johor Malaysia. School of Education Faculty of Social Sciences and Humanities Universiti Teknologi Malaysia 81310 UTM Skudai Johor Malaysia.

English is a second language in Malaysia and English language skills are vital for Malaysians. Mobile technologies such as mobile phone can be a great tool for educational purposes since it can increase students' engagement and is a popular tool for interactive digital media. This paper discusses the development of an educational mobile game called EWORD for learning English words, for primary school kids. There were four phases involved; preliminary analysis, game design, game development and game evaluation. In the first phase, the fundamental knowledge of English words and vocabulary, English language learning methods in Malaysia, traditional learning methods, and educational mobile game were studied to acquire adequate information about the requirements for educational mobile game. User Interface (UI), gameplay, and game mechanics were designed in the second phase, while the mobile game prototype was developed in the third phase. In the fourth phase, the evaluation was conducted to evaluate the usability of the educational mobile game prototype. Based on the results, the educational mobile game helps in boosting the learnability of English words among students and is capable to increase student engagement.

关键词：

Biomedical image analysis competitions: The state of current participation practice

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Eisenmann, Matthias Reinke, Annika Weru, Vivienn Tizabi, Minu Dietlinde Isensee, Fabian Adler, Tim J. Godau, Patrick Cheplygina, Veronika Kozubek, Michal Maier-Hein, Klaus Jäger, Paul F. Kopp-Schneider, Annette Maier-Hein, Lena Ali, Sharib Gupta, Anubha Kybic, Jan Noble, Alison de Solórzano, Carlos Ortiz Pachade, Samiksha Petitjean, Caroline Sage, Daniel Wei, Donglai Wilden, Elizabeth Alapatt, Deepak Andrearczyk, Vincent Baid, Ujjwal Bakas, Spyridon Balu, Niranjan Bano, Sophia Bawa, Vivek Singh Bernal, Jorge Bodenstedt, Sebastian Casella, Alessandro Choi, Jinwook Commowick, Olivier Daum, Marie Depeursinge, Adrien Dorent, Reuben Egger, Jan Eichhorn, Hannah Engelhardt, Sandy Ganz, Melanie Girard, Gabriel Hansen, Lasse Heinrich, Mattias Heller, Nicholas Hering, Alessa Huaulmé, Arnaud Kim, Hyunjeong Li, Hongwei Bran Landman, Bennett Li, Jianning Ma, Jun Martel, Anne Martín-Isla, Carlos Menze, Bjoern Nwoye, Chinedu Innocent Oreiller, Valentin Padoy, Nicolas Pati, Sarthak Payette, Kelly Sudre, Carole van Wijnen, Kimberlin Vardazaryan, Armine Vercauteren, Tom Wagner, Martin Wang, Chuanbo Yap, Moi Hoon Yu, Zeyun Yuan, Chun Zenk, Maximilian Zia, Aneeq Zimmerer, David Bao, Rina Choi, Chanyeol Cohen, Andrew Dzyubachyk, Oleh Galdran, Adrian Gan, Tianyuan Guo, Tianqi Gupta, Pradyumna Haithami, Mahmood Ho, Edward Jang, Ikbeom Li, Zhili Luo, Zhengbo Lux, Filip Makrogiannis, Sokratis Müller, Dominik Oh, Young-Tack Pang, Subeen Pape, Constantin Polat, Gorkem Reed, Charlotte Rosalie Ryu, Kanghyun Scherr, Tim Thambawita, Vajira Wang, Haoyu Wang, Xinliang Xu, Kele Yeh, Hung Yeo, Doyeob Yuan, Yixuan Zeng, Yan Zhao, Xin Abbing, Julian Adam, Jannes Adluru, Nagesh Agethen, Niklas Ahmed, Salman Al Khalil, Yasmina Alenyà, Mireia Alhoniemi, Esa An, Chengyang Arega, Tewodros Weldebirhan Avisdris, Netanell Aydogan, Dogu Baran Bai, Yingbin Calisto, Maria Baldeon Basaran, Berke Doga Beetz, Marcel Bian, Hao Blansit, Kevin Bloch, Louise Bohnsack, Robert Bosticardo, Sara Breen, Jack Brudfors, Mikael Brüngel, Raphael Cabezas, Mariano Cacciola, Alb Heidelberg Division of Intelligent Medical Systems Germany Heidelberg HI Helmholtz Imaging Germany Faculty of Mathematics and Computer Science Heidelberg University Heidelberg Germany Heidelberg Division of Biostatistics Germany Heidelberg Division of Medical Image Computing Germany Heidelberg HI Applied Vision Lab Germany IT University of Copenhagen Copenhagen Denmark Centre for Biomedical Image Analysis Masaryk University Brno Czech Republic Heidelberg Interactive Machine Learning Group Germany Faculty of Mathematics and Computer Science and Medical Faculty Heidelberg University Heidelberg Germany NCT Heidelberg DKFZ University Hospital Heidelberg Germany School of Computing University of Leeds Leeds United Kingdom SBILab Department of ECE IIIT-Delhi India Faculty of Electrical Engineering Czech Technical University Prague Czech Republic Institute of Biomedical Engineering University of Oxford United Kingdom Center for Applied Medical Research Pamplona Spain Shri Guru Gobind Singhji Institute of Engineering and Technology Maharashtra Nanded India Université de Rouen Normandie France Lausanne Switzerland School of Engineering and Applied Science Harvard University United States ICube University of Strasbourg CNRS France Institute of Informatics School of Management HES-SO Valais-Wallis University of Applied Sciences and Arts Western Switzerland Techno-Pôle 3 Sierre3960 Switzerland Department of Nuclear Medicine and Molecular Imaging Lausanne University Hospital Rue du Bugnon 46 LausanneCH-1011 Switzerland University of Pennsylvania PhiladelphiaPA United States Department of Radiology University of Washington United States Wellcome EPSRC Centre for Interventional and Surgical Sciences University College London London United Kingdom Visual Artificial Intelligence Lab Oxford Brookes University Oxford United Kingdom Universitat Autònoma de Barcelona & Computer Vision Center Spain Dresden Fetscherstraße 74 PF 64 Dresden01307 Germany

The number of international benchmarking competitions is steadily increasing in various fields of machine learning (ML) research and practice. So far, however, little is known about the common practice as well as bottlenecks faced by the community in tackling the research questions posed. To shed light on the status quo of algorithm development in the specific field of biomedical imaging analysis, we designed an international survey that was issued to all participants of challenges conducted in conjunction with the IEEE ISBI 2021 and MICCAI 2021 conferences (80 competitions in total). The survey covered participants' expertise and working environments, their chosen strategies, as well as algorithm characteristics. A median of 72% challenge participants took part in the survey. According to our results, knowledge exchange was the primary incentive (70%) for participation, while the reception of prize money played only a minor role (16%). While a median of 80 working hours was spent on method development, a large portion of participants stated that they did not have enough time for method development (32%). 25% perceived the infrastructure to be a bottleneck. Overall, 94% of all solutions were deep learning-based. Of these, 84% were based on standard architectures. 43% of the respondents reported that the data samples (e.g., images) were too large to be processed at once. This was most commonly addressed by patch-based training (69%), downsampling (37%), and solving 3D analysis tasks as a series of 2D tasks. K-fold cross-validation on the training set was performed by only 37% of the participants and only 50% of the participants performed ensembling based on multiple identical models (61%) or heterogeneous models (39%). 48% of the respondents applied postprocessing steps. © 2022, CC BY.

关键词： Benchmarking

Spatially-resolved spectro-photometric SED Modeling of NGC 253’s Central Molecular Zone I. Studying the star formation in extragalactic giant molecular clouds

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Humire, Pedro K. Dey, Subhrata Ronconi, Tommaso Sasse, Victor H. Fernandes, Roberto Cid Martín, Sergio Donevski, Darko Malek, Katarzyna Fernández-Ontiveros, Juan A. Song, Yiqing Hamed, Mahmoud Mangum, Jeffrey G. Henkel, Christian Rivilla, Víctor M. Colzi, Laura Harada, N. Demarco, Ricardo Goyal, Arti Meier, David S. Panda, Swayamtrupta Krabbe, Ângela C. Yan, Yaoting Lopes, Amanda R. Sakamoto, K. Muller, S. Tanaka, K. Yoshimura, Y. Nakanishi, K. Kanaan, Antonio Ribeiro, Tiago Schoenell, William de Oliveira, Claudia Mendes Departamento de Astronomia Instituto de Astronomia Geofísica e Ciências Atmosféricas da USP Cidade Universitária SP São Paulo05508-090 Brazil Astronomical Observatory The Jagiellonian University Orla 171 Kraków30-244 Poland National Center for Nuclear Research ul. Pasteura 7 Warsaw02-093 Poland Via Bonomea 265 TriesteIT-34136 Italy Via Beirut 2 TriesteIT-34151 Italy Via Gobetti 93/3 BolognaI-40127 Italy ICSC – Centro Nazionale di Ricerca in High Performance Computing Big Data e Quantum Computing Via Magnanelli 2 Bologna Italy Universidade Federal de Santa Catarina Campus Universitário Reitor João David Ferreira Lima SC Florianópolis88040-900 Brazil European Southern Observatory Alonso de Córdova 3107 Vitacura Santiago763-0355 Chile Joint ALMA Observatory Alonso de Córdova 3107 Vitacura Santiago763-0355 Chile Plaza San Juan 1 TeruelE–44001 Spain CSIC-INTA Carretera de Ajalvir km 4 Torrejón de Ardoz Madrid28850 Spain National Radio Astronomy Observatory 520 Edgemont Road CharlottesvilleVA22903-2475 United States Max-Planck-Institut für Radioastronomie Auf-dem-Hügel 69 Bonn53121 Germany Xinjiang Astronomical Observatory Chinese Academy of Sciences Urumqi830011 China National Astronomical Observatory of Japan 2-21-1 Osawa Mitaka Tokyo181-8588 Japan Institute of Astronomy and Astrophysics Academia Sinica 11F of AS/NTU Astronomy-Mathematics Building No.1 Sec. 4 Roosevelt Rd Taipei106319 Taiwan 2-21-1 Osawa Mitaka Tokyo181-1855 Japan Institute of Astrophysics Facultad de Ciencias Exactas Universidad Andrés Bello Sede Concepción Talcahuano Chile New Mexico Institute of Mining and Technology 801 Leroy Place SocorroNM87801 United States International Gemini Observatory NSF NOIRLab Casilla 603 La Serena Chile Instituto de Astrofísica de La Plata CONICET-UNLP Paseo del Bosque s/n B1900FWA Argentina Department of Space Earth and Environment Chalmers University of Technology Onsala Space Observatory OnsalaSE-4

Context. Studying the interstellar medium in nearby starbursts is essential for gaining insights into the physical mechanisms driving these extreme objects, thought to be analogs of young, primeval, star-forming galaxies. This task is now feasible due to deep spectro-photometric data enabled by rapid advancements in ground- and space-based facilities. To fully leverage this wealth of information, extracting insights from the spectral line properties, and the spectral energy distribution (SED) is imperative. Aims. This study aims to produce and analyze the physical properties of the first spatially-resolved multi-wavelength SED of an extragalactic source that covers six decades in frequency (from near-ultraviolet to centimeter wavelengths) at an angular resolution of 3′′ which corresponds to linear scale of ∼51 pc at the distance of NGC 253. We focus on the central molecular zone (CMZ) of this starburst galaxy, which contains giant molecular clouds (GMCs) responsible for half of the galaxy’s star formation. Methods. We retrieve archival data from optical to centimeter wavelengths, covering six decades of spectral range. We compute SEDs to fit the observations using the GalaPy code and confronting results with the CIGALE code for validation. We also employ starlight library to analyze the stellar optical spectra of the GMCs. Results. Our results reveal significant differences between central and external GMCs in terms of stellar and dust masses, star formation rates (SFRs), and bolometric luminosities, to provide a few. We have obtained tight relations between monochromatic stellar tracers and star-forming conditions obtained from panchromatic emission. We find that the best SFR tracers are radio continuum bands at 33 GHz, radio recombination lines (RRLs), and the total infrared (IR) luminosity range (LIR;8–1000µm) as well as the IR emission at 60µm. The emission line diagnostics based on the BPT and WHAN diagrams suggest that the nuclear region of NGC 253 exhibits sh

关键词： Optical emission spectroscopy

Privacy-preserving data search and sharing protocol for social networks through wireless applications

学校读者我要写书评

暂无评论

Privacy-preserving data search and sharing protocol for soci...

作者： Wang, Xiaofen Mu, Yi Chen, Rongmao Big Data Research Center and Cyber Security Research Center University of Electronic Science and Technology of China ChengduSichuan611731 China Centre for Computer and Information Security Research School of Computing and Information Technology University of Wollongong NSW2522 Australia

data search and sharing are two important functionalities in social networks. The social network users can form a peer-to-peer group and securely and flexibly search and share cloud data through wireless applications. When the number of users increases, the communication, storage, and computational overheads will be increased, and the quality of services such as searching and data sharing for clients could be affected. In order to solve these problems, we formalize an ID-based multi-user searchable encryption (IDB-MUSE) and formally define its security model, where the security notions accommodate indistinguishability against insider's keyword guessing attack, indistinguishability against chosen keyword attack, and indistinguishability against insider's identity guessing attack. We present an IDB-MUSE scheme, where the index and search trapdoor are of constant size. We formally prove its security properties. To improve the search efficiency, we divide the computation of the trapdoor into two phases, that is, the offline phase and the online phase. The computation cost for the online phase trapdoor remains constant with respect to the number of users. Based on the IDB-MUSE scheme, a privacy-preserving data search and sharing protocol is proposed, where only the authorized user can access the shared group data. It captures the properties of source authenticity, data and search pattern privacy-preserving, anonymity, and request unlinkability. The experimental results show that the protocol is practical for wireless applications. Copyright © 2016 John Wiley & Sons, Ltd. Copyright © 2016 John Wiley & Sons, Ltd.

关键词： Digital storage

A repeating fast radio burst associated with a persistent radio source

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Niu, Chen-Hui Aggarwal, K. Li, D. Zhang, X. Chatterjee, S. Tsai, C.-W. Yu, W. Law, C.J. Burke-Spolaor, S. Cordes, J.M. Zhang, Y.-K. Ocker, S.K. Yao, J.-M. Wang, P. Feng, Y. Niino, Y. Bochenek, C. Cruces, M. Connor, L. Jiang, J.-A. Dai, S. Luo, R. Li, G.-D. Miao, C.-C. Niu, J.-R. Anna-Thomas, R. Sydnor, J. Stern, D. Wang, W.-Y. Yuan, M. Yue, Y.-L. Zhou, D.-J. Yan, Z. Zhu, W.-W. Zhang, B. National Astronomical Observatories Chinese Academy of Sciences Beijing100101 China Department of Physics and Astronomy West Virginia University P.O. Box 6315 MorgantownWV26506 United States Shanghai Astronomical Observatory Chinese Academy of Sciences 80 Nandan Road Shanghai 200030 China Cornell Center for Astrophysics and Planetary Science Department of Astronomy Cornell University IthacaNY14853 United States Cahill Center for Astronomy and Astrophysics California Institute of Technology MC 249-17 PasadenaCA91125 United States Center for Gravitational Waves and Cosmology West Virginia University Chestnut Ridge Research Building MorgantownWV26505 United States Canadian Institute for Advanced Research CIFAR Azrieli Global Scholar MaRS Centre West Tower 661 University Ave. Suite 505 TorontoONM5G 1M1 Canada Owens Valley Radio Observatory California Institute of Technology 100 Leighton Lane Big Pine CA93513 United States Research Center for Intelligent Computing Platforms Zhejiang Laboratory Hangzhou311100 China Institute of Astronomy Graduate School of Science The University of Tokyo 2-21-1 Osawa Tokyo Mitaka181-0015 Japan Jet Propulsion Laboratory California Institute of Technology 4800 Oak Grove Drive M/S 264-723 PasadenaCA91109 United States University of Chinese Academy of Sciences Beijing100049 China Xinjiang Astronomical Observatory Chinese Academy of Sciences 150 Science 1-Street Xinjiang Urumqi830011 China CSIRO Space and Astronomy PO Box 76 EppingNSW1710 Australia Department of Astronomy School of Physics Peking University Beijing100871 China Department of Physics and Astronomy University of Nevada Las Vegas Las VegasNV89154 United States Max-Planck-Institut für Radioastronomie Auf dem Hügel 69 BonnD-53121 Germany The University of Tokyo 5-1-5 Kashiwanoha Chiba Kashiwa277-8583 Japan School of Science Western Sydney University Locked Bag 1797 DC South PenrithNSW2751 Australia Research Center for the

The dispersive sweep of fast radio bursts (FRBs) has been used to probe the ionized baryon content of the intergalactic medium1, which is assumed to dominate the total extragalactic dispersion. While the host galaxy contributions to dispersion measure (DM) appear to be small for most FRBs2, in at least one case there is evidence for an extreme magneto-ionic local environment3, 4 and a compact persistent radio source5. Here we report the detection and localization of the repeating FRB 20190520B, which is co-located with a compact, persistent radio source and associated with a dwarf host galaxy of high specific star formation rate at a redshift z = 0.241 ± 0.001. The estimated host galaxy DM ≈ 903+72−111 pc cm−3, nearly an order of magnitude higher than the average of FRB host galaxies2, 6, far exceeds the DM contribution of the intergalactic medium. Caution is thus warranted in inferring redshifts for FRBs without accurate host galaxy identifications. © 2021, CC BY-NC-ND.

关键词： Galaxies

Small Evolution Antenna Design Method Based on Dynamic Hill Climbing Algorithm and Orthogonal Experiment

学校读者我要写书评

暂无评论

Journal of Physics: Conference Series 2019年第1期1395卷

作者： Hui Shi Junjie Li Zhiming Xie Department of Information Engineering Shanwei Polytechnic Shanwei 516600 Guangdong China The Research Institute of Cloud Computing and Data Center Project Design Shanwei Innovation Industrial Design & Research Institute Shanwei 516600 Guangdong China The Professional Committee Cloud Computing and Big Data in Higher Vocational College Higher Education Institute in Guangdong Province Shanwei 516600 Guangdong China

Small-scale evolution antennas have the advantages of high gain, wide bandwidth and small size, and have broad application prospects in the aerospace field. Aiming at the complex high-dimensional and dynamic problems of aerial antenna modeling, a dynamic hill-climbing algorithm based on orthogonal design is designed and implemented to simulate and optimize the single-span antenna in high-frequency operation mode. In the optimization problem, this paper defines a new individual comparison mechanism, which can reflect the number of violations of constraints and the degree of violation of each constraint, making the comparison of individuals more reasonable. The antenna simulation results demonstrate the effectiveness and feasibility of using dynamic evolutionary algorithms for antenna design.

关键词：

A Face Recognition Method Based on CNN

学校读者我要写书评

暂无评论

Journal of Physics: Conference Series 2019年第1期1395卷

作者： Zhiming Xie Junjie Li Hui Shi Department of Information Engineering Shanwei Polytechnic Shanwei 516600 Guangdong China The Research Institute of Cloud Computing and Data Center Project Design Shanwei Innovation Industrial Design & Research Institute Shanwei 516600 Guangdong China The Professional Committee Cloud Computing and Big Data in Higher Vocational College Higher Education Institute in Guangdong Province Shanwei 516600 Guangdong China

The traditional face recognition technology is more complicated for the extraction of facial image features and the selection of classifiers, and the recognition rate is not high. With the continuous maturity of the convolutional neural network from handwritten digit recognition to face recognition, A face recognition algorithm that tests CNN using the Python+Keras framework. The method mainly involves two aspects. One is to observe the influence on the network by changing the number of neurons in the hidden layer; the other is to observe the influence on the network by changing the number of feature maps of the convolutional layer 1 and the convolutional layer 2. The best CNN model is 36-76-1024 through multiple sets of experimental tests. The model can automatically extract facial image features and classify them. Using adam optimizer and softmax classifier for face recognition can make training faster convergence and more. Effectively improve accuracy and use the Dropout method to avoid overfitting. The experimental results show that the recognition rate of the CNN model on the olivettifaces face database is 97.5%. When the optimal CNN model is used, the average recognition rate is close to 100%, which verifies the validity and accuracy of the algorithm and model.

关键词：