检索结果-内蒙古大学图书馆

HIGHRES: Highlight-based reference-less evaluation of summarization

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Hardy Narayan, Shashi Vlachos, Andreas Department of Computer Science University of Sheffield Google Research Department of Computer Science and Technology University of Cambridge

There has been substantial progress in summarization research enabled by the availability of novel, often large-scale, datasets and recent advances on neural network-based approaches. However, manual evaluation of the system generated summaries is inconsistent due to the difficulty the task poses to human non-expert readers. To address this issue, we propose a novel approach for manual evaluation, HIGHlight-based Reference-less Evaluation of Summarization (HIGHRES), in which summaries are assessed by multiple annotators against the source document via manually highlighted salient content in the latter. Thus summary assessment on the source document by human judges is facilitated, while the highlights can be used for evaluating multiple systems. To validate our approach we employ crowd-workers to augment with highlights a recently proposed dataset and compare two state-of-the-art systems. We demonstrate that HIGHRES improves inter-annotator agreement in comparison to using the source document directly, while they help emphasize differences among systems that would be ignored under other evaluation approaches. Copyright © 2019, The Authors. All rights reserved.

关键词： Large dataset

Kronecker-factored curvature approximations for recurrent neural networks 6

学校读者我要写书评

暂无评论

Kronecker-factored curvature approximations for recurrent ne...

6th International Conference on Learning Representations, ICLR 2018

作者： Martens, James Ba, Jimmy Johnson, Matthew DeepMind United Kingdom Department of Computer Science University of Toronto Toronto Canada Google Brain United States

Kronecker-factor Approximate Curvature (Martens & Grosse, 2015) (K-FAC) is a 2nd-order optimization method which has been shown to give state-of-the-art performance on large-scale neural network optimization tasks (Ba et al., 2017). It is based on an approximation to the Fisher information matrix (FIM) that makes assumptions about the particular structure of the network and the way it is pa-rameterized. The original K-FAC method was applicable only to fully-connected networks, although it has been recently extended by Grosse & Martens (2016) to handle convolutional networks as well. In this work we extend the method to handle RNNs by introducing a novel approximation to the FIM for RNNs. This approximation works by modelling the statistical structure between the gradient contributions at different time-steps using a chain-structured linear Gaussian graphical model, summing the various cross-moments, and computing the inverse in closed form. We demonstrate in experiments that our method significantly outperforms general purpose state-of-the-art optimizers like SGD with momentum and Adam on several challenging RNN training tasks. © Learning Representations, ICLR 2018 - Conference Track *** right reserved.

关键词： Fisher information matrix

Silly rules improve the capacity of agents to learn stable enforcement and compliance behaviors

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Köster, Raphael Hadfield-Menell, Dylan Hadfield, Gillian K. Leibo, Joel Z. DeepMind Department of Electrical Engineering and Computer Science University of California Berkeley Center for Human-Compatible AI Schwartz Reisman Institute for Technology and Society University of Toronto Vector Institute Center for Human-Compatible AI OpenAI

How can societies learn to enforce and comply with social norms? Here we investigate the learning dynamics and emergence of compliance and enforcement of social norms in a foraging game, implemented in a multi-agent reinforcement learning setting. In this spatiotemporally extended game, individuals are incentivized to implement complex berry-foraging policies and punish transgressions against social taboos covering specific berry types. We show that agents benefit when eating poisonous berries is taboo, meaning the behavior is punished by other agents, as this helps overcome a credit-assignment problem in discovering delayed health effects. Critically, however, we also show that introducing an additional taboo, which results in punishment for eating a harmless berry, improves the rate and stability with which agents learn to punish taboo violations and comply with taboos. Counterintuitively, our results show that an arbitrary taboo (a "silly rule") can enhance social learning dynamics and achieve better outcomes in the middle stages of learning. We discuss the results in the context of studying normativity as a group-level emergent phenomenon. Copyright © 2020, The Authors. All rights reserved.

关键词： Fruits

Transoceanic phase and polarization fiber sensing using real-time coherent transceiver

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Mazur, Mikael Castellanos, Jorge C. Ryf, Roland Börjeson, Erik Chodkiewicz, Tracy Kamalov, Valey Yin, Shuang Fontaine, Nicolas K. Chen, Haoshuo Dallachiesa, Lauren Corteselli, Steve Copping, Philip Gripp, Jürgen Mortelette, Aurelien Kowalski, Benoit Dellinger, Rodney Neilson, David T. Larsson-Edefors, Per Nokia Bell Labs 600 Mountain Ave. Murray HillNJ07974 United States Google LLC Mountain ViewCA94043 United States Department of Computer Science and Engineering Chalmers University of Technology Sweden Nokia 600 Mountain Ave. Murray HillNJ07974 United States

We implement a real-time coherent transceiver with fast streaming outputs for environmental sensing. Continuous sensing using phase and equalizer outputs over 12800km of a submarine cable is demonstrated to enable time resolved spectroscopy in broad spectral range of 10mHz - 1kHz. Copyright © 2021, The Authors. All rights reserved.

关键词： Laser spectroscopy

Low-field magnetic resonance image enhancement via stochastic image quality transfer

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Lin, Hongxiang Figini, Matteo D'Arco, Felice Ogbole, Godwin Tanno, Ryutaro Blumberg, Stefano B. Ronan, Lisa Brown, Biobele J. Carmichael, David W. Lagunju, Ikeoluwa Cross, Judith Helen Fernandez-Reyes, Delmiro Alexander, Daniel C. Research Center for Healthcare Data Science Zhejiang Lab Zhejiang Hangzhou311121 China Centre for Medical Image Computing University College London LondonWC1E 6BT United Kingdom Department of Computer Science University College London LondonWC1E 6BT United Kingdom Department of Radiology Great Ormond Street Hospital for Children LondonWC1N 3JH United Kingdom Department of Radiology College of Medicine University of Ibadan Ibadan200284 Nigeria Google DeepMind LondonN1C 4AG United Kingdom Centre for Artificial Intelligence University College London LondonWC1E 6BT United Kingdom Department of Paediatrics College of Medicine University of Ibadan Ibadan200284 Nigeria School of Biomedical Engineering & Imaging Sciences King’s College London LondonNW3 3ES United Kingdom UCL Great Ormond Street Institute of Child Health LondonWC1N 3JH United Kingdom

关键词： Deep neural networks

Exact and approximate hierarchical clustering using A

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Greenberg, Craig S. Macaluso, Sebastian Monath, Nicholas Dubey, Avinava Flaherty, Patrick Zaheer, Manzil Ahmed, Amr Cranmer, Kyle McCallum, Andrew National Institute of Standards and Technology United States Center for Cosmology and Particle Physics Center for Data Science New York University United States College of Information and Computer Sciences University of Massachusetts Amherst United States Google Research Mountain ViewCA United States Department of Mathematics and Statistics University of Massachusetts Amherst United States

Hierarchical clustering is a critical task in numerous domains. Many approaches are based on heuristics and the properties of the resulting clusterings are studied post hoc. However, in several applications, there is a natural cost function that can be used to characterize the quality of the clustering. In those cases, hierarchical clustering can be seen as a combinatorial optimization problem. To that end, we introduce a new approach based on A* search. We overcome the prohibitively large search space by combining A* with a novel trellis data structure. This combination results in an exact algorithm that scales beyond previous state of the art, from a search space with 1012 trees to 1015 trees, and an approximate algorithm that improves over baselines, even in enormous search spaces that contain more than 101000 trees. We empirically demonstrate that our method achieves substantially higher quality results than baselines for a particle physics use case and other clustering benchmarks. We describe how our method provides significantly improved theoretical bounds on the time and space complexity of A* for clustering. Copyright © 2021, The Authors. All rights reserved.

关键词： Cost functions

Optimal scaling quantum linear systems solver via discrete adiabatic theorem

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Costa, Pedro C.S. An, Dong Sanders, Yuval R. Su, Yuan Babbush, Ryan Berry, Dominic W. Department of Physics and Astronomy Macquarie University SydneyNSW2109 Australia Joint Center for Quantum Information and Computer Science University of Maryland College ParkMD20742 United States Google Quantum Ai VeniceCA90291 United States Centre for Quantum Software and Information University of Technology Sydney SydneyNSW2007 Australia

Recently, several approaches to solving linear systems on a quantum computer have been formulated in terms of the quantum adiabatic theorem for a continuously varying Hamiltonian. Such approaches enabled near-linear scaling in the condition number κ of the linear system, without requiring a complicated variable-time amplitude amplification procedure. However, the most efficient of those procedures is still asymptotically sub-optimal by a factor of log(κ). Here, we prove a rigorous form of the adiabatic theorem that bounds the error in terms of the spectral gap for intrinsically discrete time evolutions. We use this discrete adiabatic theorem to develop a quantum algorithm for solving linear systems that is asymptotically optimal, in the sense that the complexity is strictly linear in κ, matching a known lower bound on the complexity. Our O(κ log(1/ϵ)) complexity is also optimal in terms of the combined scaling in κ and the precision ϵ. Compared to existing suboptimal methods, our algorithm is simpler and easier to implement. Moreover, we determine the constant factors in the algorithm, which would be suitable for determining the complexity in terms of gate counts for specific applications. Copyright © 2021, The Authors. All rights reserved.

关键词： Linear systems

Gender equity in technologies: Considerations for design in the global south

学校读者我要写书评

暂无评论

Interactions 2018年第1期25卷 58-61页

作者： Sambasivan, Nithya Checkley, Garen Ahmed, Nova Batool, Amna Google United States Department of Electrical and Computer Engineering North South University Bangladesh Information Technology University Department of Computer Science Pakistan Pakistan

This is a forum for perspectives on designing for communities marginalized by economics, social status, infrastructure, or policies. It will discuss design methods, theoretical and conceptual contributions, and method...

关键词：

Training of Physical Neural Networks

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Momeni, Ali Rahmani, Babak Scellier, Benjamin Wright, Logan G. McMahon, Peter L. Wanjura, Clara C. Li, Yuhang Skalli, Anas Berloff, Natalia G. Onodera, Tatsuhiro Oguz, Ilker Morichetti, Francesco del Hougne, Philipp Le Gallo, Manuel Sebastian, Abu Mirhoseini, Azalia Zhang, Cheng Marković, Danijela Brunner, Daniel Moser, Christophe Gigan, Sylvain Marquardt, Florian Ozcan, Aydogan Grollier, Julie Liu, Andrea J. Psaltis, Demetri Alù, Andrea Fleury, Romain Lausanne Switzerland Microsoft Research 198 Cambridge Science Park CambridgeCB4 0AB United Kingdom Rain AI San Francisco United States Department of Applied Physics Yale University CT United States School of Applied and Engineering Physics Cornell University IthacaNY14853 United States Max Planck Institute for the Science of Light Staudtstraße 2 Erlangen91058 Germany Department of Electrical and Computer Engineering University of California Los AngelesCA90095 United States FEMTO-ST Institute Optics Department CNRS University Bourgogne Franche-Comté Besançon25030 Cedex France Department of Applied Mathematics and Theoretical Physics University of Cambridge Cambridge United Kingdom NTT Physics and Informatics Laboratories NTT Research Inc. Sunnyvale United States Lausanne Switzerland Dipartimento di Elettronica Informazione e Bioingegneria Politecnico di Milano Milan Italy Univ Rennes CNRS IETR UMR 6164 RennesF-35000 France IBM Research Europe– Zurich Rüschlikon8803 Switzerland Department of Computer Science Stanford University United States Google DeepMind 1600 Amphitheatre Parkway Mountain ViewCA94043 United States Unité Mixte de Physique CNRS/Thales CNRS Thales Université Paris-Saclay Palaiseau France Laboratoire Kastler Brossel Sorbonne Université École Normale Supérieure Collège de France CNRS UMR 8552 Paris France Laboratoire Albert Fert CNRS Thales UniversitéParis-Saclay Palaiseau91767 France Department of Physics and Astronomy University of Pennsylvania PhiladelphiaPA19104 United States Lausanne Switzerland Photonics Initiative Advanced Science Research Center City University of New York New YorkNY10031 United States Physics Program Graduate Center City University of New York New YorkNY10016 United States

Physical neural networks (PNNs) are a class of neural-like networks that leverage the properties of physical systems to perform computation. While PNNs are so far a niche research area with small-scale laboratory demonstrations, they are arguably one of the most underappreciated important opportunities in modern artificial intelligence (AI). Could we train AI models 1000x larger than current ones? Could we do this and also have them perform inference locally and privately on edge devices, such as smartphones or sensors? Research over the past few years has shown that the answer to all these questions is likely "textityes, with enough research": PNNs could one day radically change what is possible and practical for AI systems. To do this will however require rethinking both how AI models work, and how they are trained – primarily by considering the problems through the constraints of the underlying hardware physics. To train PNNs at large scale, many methods including backpropagation-based and backpropagation-free approaches are now being explored. These methods have various trade-offs, and so far no method has been shown to scale to the same scale and performance as the backpropagation algorithm widely used in deep learning today. However, this is rapidly changing, and a diverse ecosystem of training techniques provides clues for how PNNs may one day be utilized to create both more efficient realizations of current-scale AI models, and to enable unprecedented-scale models. Copyright © 2024, The Authors. All rights reserved.

关键词： Deep learning