检索结果-内蒙古大学图书馆

TR-BERT: Dynamic token reduction for accelerating BERT inference

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Ye, Deming Lin, Yankai Huang, Yufei Sun, Maosong Department of Computer Science and Technology Tsinghua University Beijing China Institute for Artificial Intelligence Tsinghua University Beijing China Beijing National Research Center for Information Science and Technology China State Key Lab on Intelligent Technology and Systems Tsinghua University Beijing China Beijing Academy of Artificial Intelligence China Pattern Recognition Center WeChat AI Tencent Inc

Existing pre-trained language models (PLMs) are often computationally expensive in inference, making them impractical in various resource-limited real-world applications. To address this issue, we propose a dynamic token reduction approach to accelerate PLMs' inference, named TR-BERT, which could flexibly adapt the layer number of each token in inference to avoid redundant calculation. Specially, TR-BERT formulates the token reduction process as a multi-step token selection problem and automatically learns the selection strategy via reinforcement learning. The experimental results on several downstream NLP tasks show that TR-BERT is able to speed up BERT by 2-5 times to satisfy various performance demands. Moreover, TR-BERT can also achieve better performance with less computation in a suite of long-text tasks since its token-level layer number adaption greatly accelerates the self-attention operation in PLMs. The source code and experiment details of this paper can be obtained from https://***/thunlp/TR-BERT. Copyright © 2021, The Authors. All rights reserved.

关键词： Reinforcement learning

Detach and Enhance: Learning Disentangled Cross-modal Latent Representation for Efficient Face-Voice Association and Matching

学校读者我要写书评

暂无评论

Detach and Enhance: Learning Disentangled Cross-modal Latent...

IEEE International Conference on Data Mining (ICDM)

作者： Zhenning Yu Xin Liu Yiu-Ming Cheung Minghang Zhu Xing Xu Nannan Wang Taihao Li Dept. of Comput. Sci. & Fujian Key Lab. of Big Data Intelligence and Security Huaqiao University Xiamen China Zhejiang Lab Hangzhou China Dept. of Comput. Sci. and Institute of Research and Continuing Education HK Baptist University Hong Kong SAR China Xiamen Key Lab. of Computer Vision and Pattern Recognition Huaqiao University Xiamen China Dept. of Computer Sci. and Eng. University of Electronic Science and Technology of China Chengdu China State Key Lab. of Integrated Services Networks & School of Telecommun. Eng. Xidian University Xi’an China

Many researches in cognitive science have shown that humans often perform face-voice association for various perception tasks, and some recent data mining works have been designed in emulating such ability intelligently. Nevertheless, most methods often suffer from the degraded performance when there exist semantically irrelevant interference factors across different modalities. To alleviate this concern, this paper presents an efficient Disentangled Cross-modal Latent Representation (DCLR) method to adaptively detach the discriminative feature attributes and enhance the face-voice association. To be specific, the proposed DCLR framework consists of two-stage cross-modal disentangling process. First, the former stage employs the supervised contrastive learning to push the representations of face-voice data from the same person closer while pulling those representations of different person away. Then, the latter stage freezes all the parameters of the former stage, and further innovates a multi-layer orthogonal decoupling scheme to learn the disentangled latent representations, while filtering out the modality-dependent irrelevant factors. Besides, the cross-modal reconstruction loss is further utilized to narrow down the semantic gap between heterogeneous feature expressions. Through the joint exploitation of the above, the proposed framework can well associate the face-voice data to benefit various kinds of cross-modal perception tasks. Extensive experiments verify the superiorities of the proposed face-voice association framework and show its competitive performances.

关键词： Representation learning Filtering Semantics Interference Data models Cognitive science Data mining

DCFH: A dynamic clustering approach based on fire hawk optimizer in flying ad hoc networks

学校读者我要写书评

暂无评论

Vehicular Communications 2024年 47卷

作者： Hosseinzadeh, Mehdi Ali, Saqib Ahmad, Husham Jawad Alanazi, Faisal Yousefpoor, Mohammad Sadegh Yousefpoor, Efat Darwesh, Aso Rahmani, Amir Masoud Lee, Sang-Woong Institute of Research and Development Duy Tan University Da Nang Viet Nam School of Medicine and Pharmacy Duy Tan University Da Nang Viet Nam Department of Information Systems College of Economics and Political Science Sultan Qaboos University Al Khoudh Muscat Oman Department of Communication and Computer Engineering Cihan University-Erbil Kurdistan Region Iraq Department of Electrical Engineering College of Engineering Prince Sattam bin Abdulaziz University Al-Kharj 11942 Saudi Arabia Center of Research and Strategic Studies Lebanese French University Kurdistan Region Iraq Department of Information Technology University of Human Development Kurdistan Region Sulaymaniyah Iraq Future Technology Research Center National Yunlin University of Science and Technology Yunlin Taiwan Pattern Recognition and Machine Learning Lab Gachon University 1342 Seongnamdaero Sujeonggu Seongnam 13120 South Korea

In flying ad hoc networks (FANETs), unmanned aerial vehicles (UAVs) communicate with each other without any fixed infrastructure. Because of frequent topological changes, instability of wireless communication, three-dimensional movement of UAVs, and limited resources, especially energy, FANETs deal with many challenges, especially the instability of UAV swarms. One solution to address these problems is clustering because it maintains network performance and increases scalability. In this paper, a dynamic clustering scheme based on fire hawk optimizer (DCFH) is proposed for FANETs. In DCFH, each cluster head calculates the period of hello messages in its cluster based on its velocity. Then, a fire hawk optimizer (FHO)-based dynamic clustering operation is carried out to determine the role of each UAV (cluster head (CH) or cluster member (CM)) in the network. To calculate the fitness value of each fire hawk, a fitness function is suggested based on four elements, namely the balance of energy consumption, the number of isolated clusters, the distribution of CHs, and the neighbor degree. To improve cluster stability, each CH manages the movement of its CMs and adjusts it based on its movement in the network. In the last phase, DCFH defines a greedy routing process to determine the next-hop node based on a score, which consists of distance between CHs, energy, and buffer capacity. Finally, DCFH is simulated using the network simulator version 2 (NS2), and its performance is compared with three methods, including the mobility-based weighted cluster routing scheme (MWCRSF), the dynamic clustering mechanism (DCM), and the Grey wolf optimization (GWO)-based clustering protocol. The simulation results show that DCFH well manages the number of clusters in the network. It improves the cluster construction time (about 55.51%), cluster lifetime (approximately 11.13%), energy consumption (about 15.16%), network lifetime (about 2.6%), throughput (approximately 3.9%), packet deliver

关键词： Artificial intelligence (AI) Clustering Flying ad hoc networks (FANETs) Optimization algorithms Unmanned aerial vehicles (UAVs)

Fine-Grained Topography and Modularity of the Macaque Frontal Pole Cortex Revealed by Anatomical Connectivity Profiles

学校读者我要写书评

暂无评论

Neuroscience Bulletin 2020年第12期36卷 1454-1473页

作者： Bin He Long Cao Xiaoluan Xia Baogui Zhang Dan Zhang Bo You Lingzhong Fan Tianzi Jiang School of Mechanical and Power Engineering Harbin University of Science and TechnologyHarbin 150080China Brainnetome Center Institute of AutomationChinese Academy of SciencesBeijing 100190China National Laboratory of Pattern Recognition Institute of AutomationChinese Academy of Sciences(CAS)Beijing 100190China Center for Excellence in Brain Science and Intelligence Technology Institute of AutomationCASBeijing 100190China Key Laboratory for Neuroinformation of the Ministry of Education School of Life Science and TechnologyUniversity of Electronic Science and Technology of ChinaChengdu 610054China The Queensland Brain Institute University of QueenslandBrisbaneQLD 4072Australia University of CAS Beijing 100049China College of Information and Computer Taiyuan University of TechnologyTaiyuan 030600China Chinese Institute for Brain Research Beijing 102206China Core Facility Center of Biomedical AnalysisTsinghua UniversityBeijing 100084China

The frontal pole cortex(FPC)plays key roles in various higher-order functions and is highly developed in non-human *** essential missing piece of information is the detailed anatomical connections for finer parcellation of the macaque FPC than provided by the previous tracer *** is important for understanding the functional architecture of the cerebral ***,combining cross-validation and principal component analysis,we formed a tractography-based parcellation scheme that applied a machine learning algorithm to divide the macaque FPC(2 males and 6 females)into eight subareas using high-resolution diffusion magnetic resonance imaging with the 9.4 T Bruker system,and then revealed their subregional ***,we applied improved hierarchical clustering to the obtained parcels to probe the modular structure of the subregions,and found that the dorsolateral FPC,which contains an extension to the medial FPC,was mainly connected to regions of the default-mode *** ventral FPC was mainly involved in the social-interaction network and the dorsal FPC in the metacognitive *** results enhance our understanding of the anatomy and circuitry of the macaque brain,and contribute to FPC-related clinical research.

关键词： Macaque Frontal pole cortex Anatomical connectivity profile Parcellation Neuroimaging Principal component analysis

Box-driven Class-wise Region Masking and Filling Rate Guided Loss for Weakly Supervised Semantic Segmentation

学校读者我要写书评

暂无评论

Box-driven Class-wise Region Masking and Filling Rate Guided...

IEEE/CVF Conference on computer vision and pattern recognition

作者： Chunfeng Song Yan Huang Wanli Ouyang Liang Wang Center for Research on Intelligent Perception and Computing (CRIPAC) National Laboratory of Pattern Recognition (NLPR) Institute of Automation Chinese Academy of Sciences (CASIA) The University of Sydney SenseTime Computer Vision Research Group

ISBN: (纸本)9781728132945

Semantic segmentation has achieved huge progress via adopting deep Fully Convolutional Networks (FCN). However, the performance of FCN based models severely rely on the amounts of pixel-level annotations which are expensive and time-consuming. To address this problem, it is a good choice to learn to segment with weak supervision from bounding boxes. How to make full use of the class-level and region-level supervisions from bounding boxes is the critical challenge for the weakly supervised learning task. In this paper, we first introduce a box-driven class-wise masking model (BCM) to remove irrelevant regions of each class. Moreover, based on the pixel-level segment proposal generated from the bounding box supervision, we could calculate the mean filling rates of each class to serve as an important prior cue, then we propose a filling rate guided adaptive loss (FR-Loss) to help the model ignore the wrongly labeled pixels in proposals. Unlike previous methods directly training models with the fixed individual segment proposals, our method can adjust the model learning with global statistical information. Thus it can help reduce the negative impacts from wrongly labeled proposals. We evaluate the proposed method on the challenging PASCAL VOC 2012 benchmark and compare with other methods. Extensive experimental results show that the proposed method is effective and achieves the state-of-the-art results.

关键词： bounding boxes filling rate Masking Semantics Supervision Regulated industries

Predictive coding and stochastic resonance as fundamental principles of auditory perception

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Schilling, Achim Sedley, William Gerum, Richard Metzner, Claus Tziridis, Konstantin Maier, Andreas Schulze, Holger Zeng, Fan-Gang Friston, Karl J. Krauss, Patrick Neuroscience Lab University Hospital Erlangen Germany Cognitive Computational Neuroscience Group University Erlangen-Nürnberg Germany Translational and Clinical Reseearch Institute Newcastle University Medical School Newcastle upon TyneNE2 4HH United Kingdom Department of Physics Center for Vision Research York University TorontoON Canada Biophysics Lab University Erlangen-Nürnberg Germany Pattern Recognition Lab University Erlangen-Nürnberg Germany Center for Hearing Research Department of Anatomy and Neurobiology Biomedical Engineering Cognitive Sciences Otolaryngology-Head and Neck Surgery University of California Irvine IrvineCA92697 United States Wellcome Centre for Human Neuroimaging Institute of Neurology University College London LondonWC1N 3AR United Kingdom

How is information processed in the brain during perception? Mechanistic insight is achieved only when experiments are employed to test formal or computational models. In analogy to lesion studies, phantom perception may serve as a vehicle to understand the fundamental processing principles underlying auditory perception. With a special focus on tinnitus - as the prime example of auditory phantom perception - we review recent work at the intersection of artificial intelligence, psychology, and neuroscience. In particular, we discuss why everyone with tinnitus suffers from hearing loss, but not everyone with hearing loss suffers from tinnitus. We argue that the increase of sensory precision due to Bayesian inference could be caused by intrinsic neural noise and lead to a prediction error in the cerebral cortex. Hence, two fundamental processing principles - being ubiquitous in the brain - provide the most explanatory power for the emergence of tinnitus: predictive coding as a top-down, and stochastic resonance as a complementary bottom-up mechanism. We conclude that both principles play a crucial role in healthy auditory perception. © 2022, CC BY.

关键词： Neural networks

Weakly supervised geodesic segmentation of egyptian mummy CT scans

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Hati, Avik Bustreo, Matteo Sona, Diego Murino, Vittorio Del Bue, Alessio Pattern Analysis and Computer Vision Istituto Italiano di Tecnologia Genova Italy Neuroinformatics Laboratory Fondazione Bruno Kessler Trento Italy Huawei Technologies Ltd. Ireland Research Center Dublin Ireland Department of Computer Science Università di Verona Italy

In this paper, we tackle the task of automatically analyzing 3D volumetric scans obtained from computed tomography (CT) devices. In particular, we address a particular task for which data is very limited: the segmentation of ancient Egyptian mummies CT scans. We aim at digitally unwrapping the mummy and identify different segments such as body, bandages and jewelry. The problem is complex because of the lack of annotated data for the different semantic regions to segment, thus discouraging the use of strongly supervised approaches. We, therefore, propose a weakly supervised and efficient interactive segmentation method to solve this challenging problem. After segmenting the wrapped mummy from its exterior region using histogram analysis and template matching, we first design a voxel distance measure to find an approximate solution for the body and bandage segments. Here, we use geodesic distances since voxel features as well as spatial relationship among voxels is incorporated in this measure. Next, we refine the solution using a GrabCut based segmentation together with a tracking method on the slices of the scan that assigns labels to different regions in the volume, using limited supervision in the form of scribbles drawn by the user. The efficiency of the proposed method is demonstrated using visualizations and validated through quantitative measures and qualitative unwrapping of the mummy. Copyright © 2020, The Authors. All rights reserved.

关键词： Geodesy

PIPAL: a Large-Scale Image Quality Assessment Dataset for Perceptual Image Restoration

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Gu, Jinjin Cai, Haoming Chen, Haoyu Ye, Xiaoxing Ren, Jimmy S. Dong, Chao School of Data Science Chinese University of Hong Kong Shenzhen China ShenZhen Key Lab of Computer Vision and Pattern Recognition SIAT-SenseTime Joint Lab Shenzhen Institutes of Advanced Technology Chinese Academy of Sciences China SenseTime Research SIAT Branch Shenzhen Institute of Artificial Intelligence and Robotics for Society China

Image quality assessment (IQA) is the key factor for the fast development of image restoration (IR) algorithms. The most recent IR methods based on Generative Adversarial Networks (GANs) have achieved significant improvement in visual performance, but also presented great challenges for quantitative evaluation. Notably, we observe an increasing inconsistency between perceptual quality and the evaluation results. Then we raise two questions: (1) Can existing IQA methods objectively evaluate recent IR algorithms? (2) When focus on beating current benchmarks, are we getting better IR algorithms? To answer these questions and promote the development of IQA methods, we contribute a large-scale IQA dataset, called Perceptual Image Processing Algorithms (PIPAL) dataset. Especially, this dataset includes the results of GAN-based methods, which are missing in previous datasets. We collect more than 1.13 million human judgments to assign subjective scores for PIPAL images using the more reliable "Elo system". Based on PIPAL, we present new benchmarks for both IQA and super-resolution methods. Our results indicate that existing IQA methods cannot fairly evaluate GAN-based IR algorithms. While using appropriate evaluation methods is important, IQA methods should also be updated along with the development of IR algorithms. At last, we improve the performance of IQA networks on GAN-based distortions by introducing anti-aliasing pooling. Experiments show the effectiveness of the proposed method. Copyright © 2020, The Authors. All rights reserved.

关键词： Image reconstruction

Structuring Nonlinear Wavefront Emitted from Monolayer Transition-Metal Dichalcogenides

学校读者我要写书评

暂无评论

research 2020年第1期2020卷 1721-1730页

作者： Xuanmiao Hong Guangwei Hu Wenchao Zhao Kai Wang Shang Sun Rui Zhu Jing Wu Weiwei Liu Kian Ping Loh Andrew Thye Shen Wee Bing Wang Andrea Alù Cheng-Wei Qiu Peixiang Lu Wuhan National Laboratory for Optoelectronics and School of Physics Huazhong University of Science and TechnologyWuhan 430074China Department of Electrical and Computer Engineering National University of Singapore4 Engineering Drive 3Singapore 117583 Advanced Science Research Center City University of New YorkNew York 10031USA Department of Physics National University of Singapore2 Science Drive 3Singapore 117542 Institute of Materials Research and Engineering Agency for ScienceTechnology and Research2 Fusionopolis WayInnovis#08-03Singapore 138634 Department of Chemistry National University of Singapore3 Science Drive 3Singapore 17543 Centre for Advanced 2D Materials National University of SingaporeBlock S146 Science Drive 2Singapore 117546 Hubei Key Laboratory of Optical Information and Pattern Recognition Wuhan Institute of TechnologyWuhan 430205China

The growing demand for tailored nonlinearity calls for a structure with unusual phase discontinuity that allows the realization of nonlinear optical chirality,holographic imaging,and nonlinear wavefront ***-metal dichalcogenide(TMDC)monolayers offer giant optical nonlinearity within a few-angstrom thickness,but limitations in optical absorption and domain size impose restriction on wavefront control of nonlinear emissions using classical light *** contrast,noble metal-based plasmonic nanosieves support giant field enhancements and precise nonlinear phase control,with hundred-nanometer pixellevel resolution;however,they suffer from intrinsically weak nonlinear ***,we report a multifunctional nonlinear interface by integrating TMDC monolayers with plasmonic nanosieves,yielding drastically different nonlinear functionalities that cannot be accessed by either *** a hybrid nonlinear interface allows second-harmonic(SH)orbital angular momentum(OAM)generation,beam steering,versatile polarization control,and holograms,with an effective SH nonlinearityχ^((2))of~25 nm/*** designer platform synergizes the TMDC monolayer and plasmonic nanosieves to empower tunable geometric phases and large field enhancement,paving the way toward multifunctional and ultracompact nonlinear optical devices.

关键词： nonlinearity Transition yielding