Filters








889 Hits in 1.6 sec

3DN: 3D Deformation Network [article]

Weiyue Wang, Duygu Ceylan, Radomir Mech, Ulrich Neumann
2019 arXiv   pre-print
Wang et al. [25] introduce a graph-based network to reconstruct 3D manifold shapes from input images.  ... 
arXiv:1903.03322v1 fatcat:ejqbkwnedvcbvf5khiro4hce7q

Depth-aware CNN for RGB-D Segmentation [article]

Weiyue Wang, Ulrich Neumann
2018 arXiv   pre-print
Convolutional neural networks (CNN) are limited by the lack of capability to handle geometric information due to the fixed grid kernel structure. The availability of depth data enables progress in RGB-D semantic segmentation with CNNs. State-of-the-art methods either use depth as additional images or process spatial information in 3D volumes or point clouds. These methods suffer from high computation and memory cost. To address these issues, we present Depth-aware CNN by introducing two
more » ... e, flexible and effective operations: depth-aware convolution and depth-aware average pooling. By leveraging depth similarity between pixels in the process of information propagation, geometry is seamlessly incorporated into CNN. Without introducing any additional parameters, both operators can be easily integrated into existing CNNs. Extensive experiments and ablation studies on challenging RGB-D semantic segmentation benchmarks validate the effectiveness and flexibility of our approach.
arXiv:1803.06791v1 fatcat:dxogeqi4grc6ngmxiwpvmbjk5a

On State Estimation with Bad Data Detection [article]

Weiyu Xu, Meng Wang, Ao Tang
2011 arXiv   pre-print
In this paper, we consider the problem of state estimation through observations possibly corrupted with both bad data and additive observation noises. A mixed ℓ_1 and ℓ_2 convex programming is used to separate both sparse bad data and additive noises from the observations. Through using the almost Euclidean property for a linear subspace, we derive a new performance bound for the state estimation error under sparse bad data and additive observation noises. Our main contribution is to provide
more » ... rp bounds on the almost Euclidean property of a linear subspace, using the "escape-through-a-mesh" theorem from geometric functional analysis. We also propose and numerically evaluate an iterative convex programming approach to performing bad data detections in nonlinear electrical power networks problems.
arXiv:1105.0442v1 fatcat:xifmn4glc5atflxqrloai4f5ni

Learning Efficient Convolutional Networks through Irregular Convolutional Kernels [article]

Weiyu Guo, Jiabin Ma, Liang Wang, Yongzhen Huang
2019 arXiv   pre-print
As deep neural networks are increasingly used in applications suited for low-power devices, a fundamental dilemma becomes apparent: the trend is to grow models to absorb increasing data that gives rise to memory intensive; however low-power devices are designed with very limited memory that can not store large models. Parameters pruning is critical for deep model deployment on low-power devices. Existing efforts mainly focus on designing highly efficient structures or pruning redundant
more » ... ns for networks. They are usually sensitive to the tasks or relay on dedicated and expensive hashing storage strategies. In this work, we introduce a novel approach for achieving a lightweight model from the views of reconstructing the structure of convolutional kernels and efficient storage. Our approach transforms a traditional square convolution kernel to line segments, and automatically learn a proper strategy for equipping these line segments to model diverse features. The experimental results indicate that our approach can massively reduce the number of parameters (pruned 69% on DenseNet-40) and calculations (pruned 59% on DenseNet-40) while maintaining acceptable performance (only lose less than 2% accuracy).
arXiv:1909.13239v1 fatcat:bcvbvuawrbeuxe6djykhbqxraq

The Limits of Error Correction with lp Decoding [article]

Meng Wang and Weiyu Xu and Ao Tang
2010 arXiv   pre-print
An unknown vector f in R^n can be recovered from corrupted measurements y = Af + e where A^(m*n)(m>n) is the coding matrix if the unknown error vector e is sparse. We investigate the relationship of the fraction of errors and the recovering ability of lp-minimization (0 < p <= 1) which returns a vector x minimizing the "lp-norm" of y - Ax. We give sharp thresholds of the fraction of errors that determine the successful recovery of f. If e is an arbitrary unknown vector, the threshold strictly
more » ... creases from 0.5 to 0.239 as p increases from 0 to 1. If e has fixed support and fixed signs on the support, the threshold is 2/3 for all p in (0, 1), while the threshold is 1 for l1-minimization.
arXiv:1006.0277v1 fatcat:3mraj74sxzbejncuexdmaxsjse

Impartial SWIPT-Assisted User Cooperation Schemes [article]

Weiyu Chen, Haiyang Ding, Shilian Wang, Daniel Benevides da Costa and Fengkui Gong
2019 arXiv   pre-print
In this paper, we propose an impartial simultaneous wireless information and power transfer (SWIPT)-assisted cooperation mechanism for a non-orthogonal multiple access (NOMA) downlink scenario. Specifically, both a cell-center user and a cell-edge user apply the power-splitting technique and utilize the harvested energy to forward the other user's information on the premise of successful decoding of their own information. Both analytical and numerical results show that the proposed impartial
more » ... r cooperation mechanism (IUCM) outperforms the traditional partial cooperation mechanism in terms of outage probability, diversity order and diversity-multiplexing trade-off (DMT). For comparison, we further incorporate the IUCM into an orthogonal frequency-division multiple access (OFDMA) framework, which is shown to preserve the same diversity order, while has a worse but more flexible DMT performance in comparison with the IUCM in the NOMA framework. Although the IUCM in OFDMA has a worse outage performance, it is proved that it has the same optimal system outage probability with the IUCM in NOMA when the relaying channel between the two users is error-free.
arXiv:1910.00244v1 fatcat:pyjckdoruvaqra7dmjvwkrjgy4

Trusting Artificial Intelligence in Healthcare

Weiyu Wang, Keng Siau
2018 Americas Conference on Information Systems  
Image design, predictability, usability, and privacy are important factors that affect trust building (Siau and Wang 2018) .  ...  Research has shown that trust is crucial in organizational relationships, e-commerce, online environment, and human-technology interactions (Siau and Wang 2018) .  ... 
dblp:conf/amcis/WangS18b fatcat:xnuw7fnxdzdnpcpxn2kpbnl6qa

Sparse Recovery with Graph Constraints [article]

Meng Wang, Weiyu Xu, Enrique Mallada, Ao Tang
2013 arXiv   pre-print
Wang, E. Mallada and A. Tang are with Cornell University, Ithaca, NY. W. Xu is with the University of Iowa, Iowa City, IA. Partial and preliminary results have appeared in [34] . entries.  ... 
arXiv:1207.2829v2 fatcat:f43csrcc4zadtk3f6n4gu7al34

Mechanical Behavior of Irregular Fibers

Weiyu He, Xungai Wang
2002 Textile research journal  
Fiber buckling behavior is associated with fabric-evoked prickle, which affects the clothing comfort and aesthetics. In this paper, the flexural buckling behavior of irregular or nonuniform fibers is studied, using the finite element method (FEM). Fiber dimensional irregularities are simulated with sine waves of different magnitude, frequency and initial phase. The critical buckling loads of the simulated fibers are then calculated from the FE model. The results indicate the increasing the
more » ... of irregularity will decrease the critical buckling load of fibers, but the effect of frequency and initial phase of irregularity on fiber buckling behavior is complicated and is affected by the fiber diameter and effective length.
doi:10.1177/004051750207200703 fatcat:3a2xozkkevg3tkslxr56offkgy

Superatom-assembly induced transition from insulator to semiconductor: A theoretical study

Jia Wang, Wanrun Jiang, Weiyu Xie, Jianpeng Wang, Zhigang Wang
2018 Science China Materials  
Wang Z also acknowledges the High Performance Computing Center of Jilin Uni-versity.  ...  Author contributions Wang Z proposed the project; Wang J calculated and analyzed the results. All authors contributed to the general discussion.  ... 
doi:10.1007/s40843-018-9329-8 fatcat:zzfkvkjldveldivp4g4zh75edi

Ethical and Moral Issues with AI

Weiyu Wang, Keng Siau
2018 Americas Conference on Information Systems  
For instance, AI may cause mass unemployment, make decisions that people cannot understand and control, lead to the wealth redistribution, and replace humans eventually (Siau and Wang, 2018) .  ... 
dblp:conf/amcis/WangS18 fatcat:2ti47qmkufajtoxijli4hy7tde

Sparse Recovery with Graph Constraints: Fundamental Limits and Measurement Construction [article]

Meng Wang, Weiyu Xu, Enrique Mallada, Ao Tang
2011 arXiv   pre-print
This paper addresses the problem of sparse recovery with graph constraints in the sense that we can take additive measurements over nodes only if they induce a connected subgraph. We provide explicit measurement constructions for several special graphs. A general measurement construction algorithm is also proposed and evaluated. For any given graph G with n nodes, we derive order optimal upper bounds of the minimum number of measurements needed to recover any k-sparse vector over G (M^G_k,n).
more » ... r study suggests that M^G_k,n may serve as a graph connectivity metric.
arXiv:1108.0443v1 fatcat:7czligglhffj3bzyoub2mstxbq

Backscatter Cooperation in NOMA Communications Systems [article]

Weiyu Chen, Haiyang Ding, Shilian Wang, Daniel Benevides da Costa, Fengkui Gong, Pedro Henrique Juliano Nardelli
2020 arXiv   pre-print
In this paper, a backscatter cooperation (BC) scheme is proposed for non-orthogonal multiple access (NOMA) downlink transmission. The key idea is to enable one user to split and then backscatter part of its received signals to improve the reception at another user. To evaluate the performance of the proposed BC-NOMA scheme, three benchmark schemes are introduced. They are the non-cooperation (NC)-NOMA scheme, the conventional relaying (CR)-NOMA scheme, and the incremental relaying (IR)-NOMA
more » ... me. For all these schemes, the analytical expressions of the minimum total power to avoid information outage are derived, based on which their respective outage performance, expected rates, and diversity-multiplexing trade-off (DMT) are investigated. Analytical results show that the proposed BC-NOMA scheme strictly outperforms the NC-NOMA scheme in terms of all the three metrics. Furthermore, theoretical analyses are validated via Monte-Carlo simulations. It is shown that unlike the CR-NOMA scheme and the IR-NOMA scheme, the proposed BC-NOMA scheme can enhance the transmission reliability without impairing the transmission rate, which makes backscattering an appealing solution to cooperative NOMA downlinks.
arXiv:2006.13646v1 fatcat:a2j7pg6rhjdklcro3jj6dffvwq

Analysis of Positional Encodings for Neural Machine Translation

Jan Rosendahl, Viet Anh Khoa Tran, Weiyue Wang, Hermann Ney
2019 Zenodo  
In this work we analyze and compare the behavior of the Transformer architecture when using different positional encoding methods. While absolute and relative positional encoding perform equally strong overall, we show that relative positional encoding is vastly superior (4.4% to 11.9% BLEU) when translating a sentence that is longer than any observed training sentence. We further propose and analyze variations of relative positional encoding and observe that the number of trainable parameters
more » ... an be reduced without a performance loss, by using fixed encoding vectors or by removing some of the positional encoding vectors.
doi:10.5281/zenodo.3525024 fatcat:ka7546s3t5ccra27dp7rsbfakq

Depth-Aware CNN for RGB-D Segmentation [chapter]

Weiyue Wang, Ulrich Neumann
2018 Lecture Notes in Computer Science  
Convolutional neural networks (CNN) are limited by the lack of capability to handle geometric information due to the fixed grid kernel structure. The availability of depth data enables progress in RGB-D semantic segmentation with CNNs. State-of-the-art methods either use depth as additional images or process spatial information in 3D volumes or point clouds. These methods suffer from high computation and memory cost. To address these issues, we present Depth-aware CNN by introducing two
more » ... e, flexible and effective operations: depth-aware convolution and depth-aware average pooling. By leveraging depth similarity between pixels in the process of information propagation, geometry is seamlessly incorporated into CNN. Without introducing any additional parameters, both operators can be easily integrated into existing CNNs. Extensive experiments and ablation studies on challenging RGB-D semantic segmentation benchmarks validate the effectiveness and flexibility of our approach.
doi:10.1007/978-3-030-01252-6_9 fatcat:ecvunnbzqvb3xj7xk7sbmaxgmu
« Previous Showing results 1 — 15 out of 889 results