检索结果-内蒙古大学图书馆

Cleaning our own dust: Simulating and separating galactic dust foregrounds with neural networks

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Aylor, K. Haq, M. Knox, L. Hezaveh, Y. Perreault-Levasseur, L. Department of Physics University of California DavisCA95616 United States Department of Mathematics University of Texas DallasTX75080 United States Département de Physique Université de Montréal MontrealQCH3T 1J4 Canada Center for Computational Astrophysics Flatiron Institute New YorkNY10010 United States Montreal Institute for Learning Algorithms Université de Montréal MontrealQCH2S 3H1 Canada

Separating galactic foreground emission from maps of the cosmic microwave background (CMB), and quantifying the uncertainty in the CMB maps due to errors in foreground separation are important for avoiding biases in scientific conclusions. Our ability to quantify such uncertainty is limited by our lack of a model for the statistical distribution of the foreground emission. Here we use a Deep Convolutional Generative Adversarial Network (DCGAN) to create an effective non-Gaussian statistical model for intensity of emission by interstellar dust. For training data we use a set of dust maps inferred from observations by the Planck satellite. A DCGAN is uniquely suited for such unsupervised learning tasks as it can learn to model a complex non-Gaussian distribution directly from examples. We then use these simulations to train a second neural network to estimate the underlying CMB signal from dust-contaminated maps. We discuss other potential uses for the trained DCGAN, and the generalization to polarized emission from both dust and synchrotron. Copyright © 2019, The Authors. All rights reserved.

关键词： Dust

Kutralnet: A portable deep learning model for fire recognition

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Ayala, Angel Fernandes, Bruno Cruz, Francisco Macêdo, David Oliveira, Adriano L.I. Zanchettin, Cleber Escola Politécnica de Pernambuco Universidade de Pernambuco Recife Brazil School of Information Technology Deakin University Geelong Australia Escuela de Ingeniería Universidad Central de Chile Santiago Chile Centro de Informática Universidade Federal de Pernambuco Recife Brazil Montreal Institute for Learning Algorithms University of Montreal QC Canada Department of Chemical and Biological Engineering Northwestern University Evanston United States

Most of the automatic fire alarm systems detect the fire presence through sensors like thermal, smoke, or flame. One of the new approaches to the problem is the use of images to perform the detection. The image approach is promising since it does not need specific sensors and can be easily embedded in different devices. However, besides the high performance, the computational cost of the used deep learning methods is a challenge to their deployment in portable devices. In this work, we propose a new deep learning architecture that requires fewer floating-point operations (flops) for fire recognition. Additionally, we propose a portable approach for fire recognition and the use of modern techniques such as inverted residual block, convolutions like depth-wise, and octave, to reduce the model’s computational cost. The experiments show that our model keeps high accuracy while substantially reducing the number of parameters and flops. One of our models presents 71% fewer parameters than FireNet, while still presenting competitive accuracy and AUROC performance. The proposed methods are evaluated on FireNet and FiSmo datasets. The obtained results are promising for the implementation of the model in a mobile device, considering the reduced number of flops and parameters acquired. Copyright © 2020, The Authors. All rights reserved.

关键词： Smoke

Multi-cue adaptive emotion recognition network

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Costa, Willams Macêdo, David Zanchettin, Cleber Figueiredo, Lucas Teichrieb, Veronicca Voxar Labs Centro de Informática Universidade Federal de Pernambuco Recife Brazil Centro de Informática Universidade Federal de Pernambuco Recife Brazil Montreal Institute for Learning Algorithms University of Montreal QC Canada Department of Chemical and Biological Engineering Northwestern University Evanston United States Unidade Acadêmica de Belo Jardim Universidade Federal Rural de Pernambuco Belo Jardim Brazil

Expressing and identifying emotions through facial and physical expressions is a significant part of social interaction. The computer task for identifying emotions and allowing a more natural interaction between humans and machines is emotion recognition. The common approaches for emotion recognition focus on analyzing facial expressions and requires the automatic localization of the face in the evaluated scene. Although these approaches can classify emotion in controlled scenarios, such techniques are limited when dealing with unconstrained daily interactions. We propose a new direction for emotion recognition based on adaptive multi-cues using face and information from context and body poses. We evaluate the proposed approach in the CAER-S dataset, considering different components in a pipeline that is 21.49% better than the current state-of-the-art method. © 2021, CC BY-NC-ND.

关键词： Speech recognition

KutralNet: A Portable Deep learning Model for Fire Recognition

学校读者我要写书评

暂无评论

KutralNet: A Portable Deep Learning Model for Fire Recogniti...

International Joint Conference on Neural Networks (IJCNN)

作者： Angel Ayala Bruno Fernandes Francisco Cruz David Macêdo Adriano L. I. Oliveira Cleber Zanchettin Escola Politécnica de Pernambuco Universidade de Pernambuco Recife Brasil School of Information Technology Deakin University Geelong Australia Escuela de Ingeniería Universidad Central de Chile Santiago Chile Centro de Informática Universidade Federal de Pernambuco Recife Brasil Montreal Institute for Learning Algorithms University of Montreal Quebec Canada Department of Chemical and Biological Engineering Northwestern University Evanston United States of America

ISBN: (数字)9781728169262

ISBN: (纸本)9781728169279

Most of the automatic fire alarm systems detect the fire presence through sensors like thermal, smoke, or flame. One of the new approaches to the problem is the use of images to perform the detection. The image approach is promising since it does not need specific sensors and can be easily embedded in different devices. However, besides the high performance, the computational cost of the used deep learning methods is a challenge to their deployment in portable devices. In this work, we propose a new deep learning architecture that requires fewer floating-point operations (flops) for fire recognition. Additionally, we propose a portable approach for fire recognition and the use of modern techniques such as inverted residual block, convolutions like depth-wise, and octave, to reduce the model's computational cost. The experiments show that our model keeps high accuracy while substantially reducing the number of parameters and flops. One of our models presents 71% fewer parameters than FireNet, while still presenting competitive accuracy and AUROC performance. The proposed methods are evaluated on FireNet and FiSmo datasets. The obtained results are promising for the implementation of the model in a mobile device, considering the reduced number of flops and parameters acquired.

关键词： Computational modeling Convolution Machine learning Sensors Computational efficiency Performance evaluation Mobile handsets

Bridging the Gap: Generalising State-of-the-Art U-Net Models to Sub-Saharan African Populations

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Amod, Alyssa R. Smith, Alexandra Joubert, Pearly Raymond, Confidence Zhang, Dong Anazodo, Udunna C. Motchon, Dodzi Mutsvangwa, Tinashe E.M. Quetin, Sébastien Brain Behaviour Unit Neuroscience Institute University of Cape Town Cape Town South Africa ShockLab Dept. of Mathematics & Applied Mathematics UCT CT South Africa Applied Mathematics Division Stellenbosch University Stellenbosch South Africa Division of Biomedical Engineering Neuroscience Institute UCT CT South Africa Medical Artificial Intelligence Lab Lagos Nigeria Lawson Health Research Institute LondonON Canada Dept. of Electrical & Computer Engineering University of British Columbia Canada Montreal Neurological Institute McGill University Montréal Canada Dept. of Medicine Dept. Clinical & Radiation Oncology University of Cape Town Cape Town South Africa Medical Physics Unit Dept. of Oncology McGill University Montréal Canada Montreal Institute for Learning Algorithms Montréal Canada

A critical challenge for tumour segmentation models is the ability to adapt to diverse clinical settings, particularly when applied to poor quality neuroimaging data. The uncertainty surrounding this adaptation stems from the lack of representative datasets, leaving top-performing models without expo-sure to common artefacts found in MRI data throughout Sub-Saharan Africa (SSA). We replicated a framework that secured the 2nd position in the 2022 BraTS competition to investigate the impact of dataset composition on mod-el performance and pursued four distinct approaches through training a model with: BraTS-Africa data only (train_SSA, N=60), 2) BraTS-Adult Glioma data only (train_GLI, N=1251), 3) both datasets together (train_ALL, N=1311), and 4) through further training the train_GLI model with BraTS-Africa data (train_ftSSA). Notably, training on a smaller low-quality dataset alone (train_SSA) yielded subpar results, and training on a larger high-quality dataset alone (train_GLI) struggled to delineate oedematous tissue in the low-quality validation set. The most promising approach (train_ftSSA) involved pre-training a model on high-quality neuroimages and then fine-tuning it on the smaller low-quality dataset. This approach outperformed the others, ranking 2nd in the MICCAI BraTS Africa global challenge external testing phase. These findings underscore the significance of larger sample sizes and broad exposure to data in improving segmentation performance. Furthermore, we demonstrated there is potential for improving such models by fine-tuning them with a wider range of data locally. © 2023, CC BY.

关键词： Neuroimaging

A Fast Multiple Cue Fusing Approach for Human Emotion Recognition

学校读者我要写书评

暂无评论

SSRN

SSRN 2022年

作者： Costa, Willams Macêdo, David Zanchettin, Cleber Talavera, Estefanía Figueiredo, Lucas Teichrieb, Veronica Voxar Labs Centro de Informática Universidade Federal de Pernambuco Av. Jornalista Aníbal Fernandes s/n Recife50740-560 Brazil Centro de Informática Universidade Federal de Pernambuco Av. Jornalista Aníbal Fernandes s/n Recife50740-560 Brazil Montreal Institute for Learning Algorithms University of Montreal 6666 St-Urbain #200 MontrealH2S 3H1 Canada Department of Chemical and Biological Engineering Nortwestern University 2145 Sheridan Road EvanstonIL60208 United States Data Management and Biometrics Group University of Twente Drienerlolaan 5 Enschede7522 NB Netherlands Unidade Acadêmica de Belo Jardim Universidade Federal Rural de Pernambuco Belo JardimPE-166 100 Brazil

Emotion recognition is the task of inferring emotion through cues that can be linked to emotion, mood, and thought. Recognizing emotions allows intelligent systems to develop a deeper understanding of the behavior of the user, unveiling intentions allied to actions and then change their responses accordingly. In this work, we propose a deep framework for emotion recognition that combines multiple visual cues to understand the emotion depicted by humans in images or videos. We experimented and evaluated our proposal on the well-known CAER-S dataset and achieved competitive results against the state-of-the-art, with an accuracy of 89.76% while having an inference time at least three times faster. Furthermore, by using different visual cues, we are able to gain a broader understanding of the context that led to the particular prediction, minimizing ambiguity and allowing in-the-wild applications in which signals occluded by the environment can be compensated. © 2022, The Authors. All rights reserved.

关键词： Emotion Recognition

Fundamental limits of decentralized data shuffling

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Wan, Kai Tuninetti, Daniela Ji, Mingyue Caire, Giuseppe Piantanida, Pablo Electrical Engineering and Computer Science Department Technische Universität Berlin Berlin10587 Germany Electrical and Computer Engineering Department University of Illinois ChicagoIL60607 United States Electrical and Computer Engineering Department University of Utah Salt Lake CityUT84112 United States CentraleSupélec-French National Center for Scientific Research Université Paris-Sud 3 Rue Joliot-Curie Gif-sur-YvetteF-91192 France Montreal Institute for Learning Algorithms Université de Montreál 2920 Chemin de la Tour MontreálQCH3T1N8 Canada

Data shuffling of training data among different computing nodes (workers) has been identified as a core element to improve the statistical performance of modern large scale machine learning algorithms. Data shuffling is often considered as one of the most significant bottlenecks in such systems due to the heavy communication load. Under a master-worker architecture (where a master has access to the entire dataset and only communication between the master and the workers is allowed) coding has been recently proved to considerably reduce the communication load. This work considers a different communication paradigm referred to as decentralized data shuffling, where workers are allowed to communicate with one another via a shared link. The decentralized data shuffling problem has two phases: Workers communicate with each other during the data shuffling phase, and then workers update their stored content during the storage phase. For the case of uncoded storage (i.e., each worker directly stores a subset of bits of the dataset), this paper proposes converse and achievable bounds that are to within a factor of 3=2 of one another. The proposed schemes are also exactly optimal under the constraint of uncoded storage for either large memory size or at most four workers in the system. As a by-product, a novel distributed clique-covering scheme is proposed for distributed broadcast with side information-a setting that includes as special cases decentralized data shuffling, distributed index coding, and device-to-device coded caching. Copyright © 2018, The Authors. All rights reserved.

关键词： Machine learning

The importance of transparency and reproducibility in artificial intelligence research

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Haibe-Kains, Benjamin Adam, George Alexandru Hosny, Ahmed Khodakarami, Farnoosh Waldron, Levi Wang, Bo McIntosh, Chris Kundaje, Anshul Greene, Casey S. Hoffman, Michael M. Leek, Jeffrey T. Huber, Wolfgang Brazma, Alvis Pineau, Joelle Tibshirani, Robert Hastie, Trevor Ioannidis, John P.A. Quackenbush, John Aerts, Hugo J.W.L. Princess Margaret Cancer Centre University Health Network TorontoON Canada Department of Medical Biophysics University of Toronto TorontoON Canada Department of Computer Science University of Toronto TorontoON Canada Ontario Institute for Cancer Research TorontoON Canada Vector Institute for Artificial Intelligence TorontoON Canada Program Brigham and Women's Hospital Harvard Medical School BostonMA United States Radiation Oncology and Radiology Dana-Farber Cancer Institute Brigham and Women's Department of Epidemiology and Biostatistics Institute for Implementation Science in Population Health CUNY Graduate School of Public Health and Health Policy New YorkNY United States Peter Munk Cardiac Centre University Health Network TorontoON Canada Hospital Harvard Medical School BostonMA United States Department of Genetics Stanford University School of Medicine StanfordCA United States Dept. of Systems Pharmacology and Translational Therapeutics Perelman School of Medicine University of Pennsylvania PhiladelphiaPA United States Childhood Cancer Data Lab Alex's Lemonade Stand Foundation PhiladelphiaPA United States Department of Biostatistics Johns Hopkins Bloomberg School of Public Health BaltimoreMD United States European Molecular Biology Laboratory Genome Biology Unit Heidelberg Germany European Molecular Biology Laboratory European Bioinformatics Institute EMBL-EBI Hinxton United Kingdom McGill University MontrealQC Canada Montreal Institute for Learning Algorithms QC Canada StanfordCA United States Department of Biomedical Data Science Stanford University School of Medicine StanfordCA United States Departments of Medicine of Health Research and Policy and of Biomedical Data Science Stanford University School of Medicine StanfordCA United States Department of Statistics Stanford University School of Humanities and Sciences StanfordCA United States Department of Biostatistics Harvard T.H Chan School of Public Health BostonMA Un

In their study, McKinney et al. showed the high potential of artificial intelligence for breast cancer screening. However, the lack of detailed methods and computer code undermines its scientific value. We identify obstacles hindering transparent and reproducible AI research as faced by McKinney et al and provide solutions with implications for the broader field. Copyright © 2020, The Authors. All rights reserved.

关键词： Artificial intelligence