3,935 Hits in 1.7 sec

Just-in-Time Dynamic-Batching [article]

Sheng Zha, Ziheng Jiang, Haibin Lin, Zhi Zhang
2019 arXiv   pre-print
Batching is an essential technique to improve computation efficiency in deep learning frameworks. While batch processing for models with static feed-forward computation graphs is straightforward to implement, batching for dynamic computation graphs such as syntax trees or social network graphs is challenging due to variable computation graph structure across samples. Through simulation and analysis of a Tree-LSTM model, we show the key trade-off between graph analysis time and batching
more » ... ness in dynamic batching. Based on this finding, we propose a dynamic batching method as an extension to MXNet Gluon's just-in-time compilation (JIT) framework. We show empirically that our method yields up to 6.25 times speed-up on a common dynamic workload, a tree-LSTM model for the semantic relatedness task.
arXiv:1904.07421v1 fatcat:amclnkesfvdvfi4dnhqmzqkxe4

A Unified Convex Surrogate for the Schatten-p Norm [article]

Chen Xu, Zhouchen Lin, Hongbin Zha
2016 arXiv   pre-print
., matrix completion (Candès and Recht, 2009) , low rank representation (Liu, Lin, and Yu, 2010) , multi-task learning (Dudik, Harchaoui, and Malick, 2012) , etc.  ... 
arXiv:1611.08372v1 fatcat:cemosmekbjh2depwninhhobsou

Diagnosis and treatment of the blepharochalasis

Zai-Lin Xiao, Yuan-Ping Zhang, Xue-Ying Zhao, Xu Zha, Guo-Jiu Wu, Lin-Kun Ma
2019 Guoji Yanke Zazhi  
Blepharochalasis is a rare disease that often onsets in childhood or adolescence. It is characterized by non-painful, non-erythemaous eyelid angioedema with intermittent recurrent episodes, resulting in relaxation, atrophy and thinned of eyelid structures. The late complications, attributing to continuous underlying structural destruction, including ptosis, orbital fat prolapse, lacrimal glands prolapse and blepharophimosis etc. The pathogenesis is not clear yet, researchers have indicated that
more » ... it might be associated with inheritance and endocrine factors. Nevertheless, histopathology examination showed elastic fibers breakdown, IgA deposits and inflammatory cell infiltration. Based on this, it is assumed that the immunologic mechanism and inflammatory reaction play a significant role in the development of the disease. According to the clinical stage, there are two treatment patterns: oral medicine and surgery at present.
doi:10.3980/j.issn.1672-5123.2019.4.16 doaj:26fd0ac506344cc8ab81d3496b66d3b0 fatcat:4utcyzqg7fctbnznoqcopk2ibm

Relaxed Majorization-Minimization for Non-smooth and Non-convex Optimization [article]

Chen Xu, Zhouchen Lin, Zhenyu Zhao, Hongbin Zha
2015 arXiv   pre-print
This can be easily done by using the linearized alternating direction method with parallel splitting and adaptive penalty (LADMPSAP) (Lin, Liu, and Li, 2013) .  ...  to add the proximal term x − x k 2 2 , which has been widely used, e.g. in proximal splitting algorithm (Attouch, Bolte, and Svaiter, 2013) and alternating direction method of multipliers (ADMM) (Lin  ... 
arXiv:1511.08062v1 fatcat:dguxjxcxqjc7dnbbr6p7jjmruq

Effect of Particle Concentration on Ablation Behavior of Carbon/Carbon Composites

ZHA Bai-lin, LIN Hao, GAO Shuang-lin, LUO Lei, ZHANG Bo-wen, ZHU Jie-tang, SUN Zhen-sheng
2016 Journal of Materials Engineering  
A self-designed oxygen-kerosene ablation system was employed to study the ablation mechanism and performance of the carbon-carbon composites. Particle concentrations of the gas-solid two-phase ablation flow were 0, 1.37%, 2.22%, and 2.64% respectively. The microstructure of post-test samples was analysed through the scanning electron microscope (SEM), and the ablation rate was calculated. The influence principle of particle concentrations on the ablation rate of the carbon-carbon composites was
more » ... studied, and the mechanism of ablation was analysed. Experiment results show that with no particles involved the mass ablation rate is 0.159g/s and the linear ablation rate is 0.175mm/s, while with particles introduced, the minimum mass ablation rate and linear ablation rate are 0.432g/s and 0.843mm/s respectively, with the increase of particle concentrations, the ablation rate becomes accelerated. Particle erosion makes the ablation of sample seriously intensified, and the ablation gradient of radial fiber on erosion surface is increased with the increase of particle concentrations as well.
doi:10.11868/j.issn.1001-4381.2016.07.016 doaj:2cf0cbc6ca2e49e6bf9eb86a07d17490 fatcat:sjia5ouh7bgbjpk2q2a6efjike

DC Biasing Treatment Scheme for Zha-Qing ±800kV Project Zhalute Power Grid

Yan-bing ZHANG, Hao-tian CUI, Wen-lin LIU, Yong HUANG, Lian-guang LIU, Qing WANG
2018 DEStech Transactions on Environment Energy and Earth Science  
Combining with the construction of Zha-Qing (±800kV) project and using the historical magnetotelluric data, the three-dimensional resistivity model and the magnetic bias current model of Zhalute power  ...  Considering the influence of these factors and with the aid of GIC standard model, this paper establishes the DCBC model of Zhalute Power Grid of Zha-Qing Project.  ...  wind power of the Zhalute Power Grid transported by the Zha-Qing Project comes from the four provinces of Eastern Mongolia and Liaoning, Jilin and Heilongjiang.  ... 
doi:10.12783/dteees/epeee2018/26478 fatcat:vy2afp2s6rcp5hmgdupin7jqee

Low Rank Global Geometric Consistency for Partial-Duplicate Image Search

Li Yang, Yang Lin, Zhouchen Lin, Hongbin Zha
2014 2014 22nd International Conference on Pattern Recognition  
All existing feature point based partial-duplicate image retrieval systems are confronted with the false feature point matching problem. To resolve this issue, geometric contexts are widely used to verify the geometric consistency in order to remove false matches. However, most of the existing methods focus on local geometric contexts rather than global. Seeking global contexts has attracted a lot of attention in recent years. This paper introduces a novel global geometric consistency, based on
more » ... the low rankness of squared distance matrices of feature points, to detect false matches. We cast the problem of detecting false matches as a problem of decomposing a squared distance matrix into a low rank matrix, which models the global geometric consistency, and a sparse matrix, which models the mismatched feature points. So we arrive at a model of Robust Principal Component Analysis. Our Low Rank Global Geometric Consistency (LRGGC) is simple yet effective and theoretically sound. Extensive experimental results show that our LRGGC is much more accurate than state of the art geometric verification methods in detecting false matches and is robust to all kinds of similarity transformation (scaling, rotation, and translation) and even slight change in 3D views. Its speed is also highly competitive even compared with local geometric consistency based methods.
doi:10.1109/icpr.2014.675 dblp:conf/icpr/YangLLZ14 fatcat:ama5gkwgyraydbt5rtrw6hyg6q

Alternating Multi-bit Quantization for Recurrent Neural Networks [article]

Chen Xu, Jianqiang Yao, Zhouchen Lin, Wenwu Ou, Yuanbin Cao, Zhirong Wang, Hongbin Zha
2018 arXiv   pre-print
Zhouchen Lin is supported by National Basic Research Program of China (973 Program  ... 
arXiv:1802.00150v1 fatcat:3qpnjetxn5hkvdk2uqpnpagagq

Effects of Ablation Angle on the Ablation Behavior of C/C Composites

ZHA Bai-lin, GAO Shuang-lin, LIN Hao, LUO Lei, ZHANG Bo-wen, ZHU Jie-tang, SUN Zhen-sheng
2017 Journal of Materials Engineering  
There is a significant impact of ablation angle on the anti-ablation performance of C/C composites. A self-designed oxygen-kerosene ablation system was employed to study the ablation behavior and mechanism of C/C composites. Typical ablation angle of the experiment was 90°, 60°, 45° respectively, and particle concentration of the gas-solid two-phase ablation flow was 1.37%. The microstructure of post-test samples was observed with the scanning electron microscope(SEM), and the ablation rate of
more » ... amples was also calculated. The result shows that when the ablation angle is 90°, 60°, 45°,the corresponding mass ablation rate for ablation experiment is 0.146,0.123,0.100g/s respectively. With the decrease of ablation angle, the mass ablation rate of samples is decreased gradually; while for the samples affected by particle erosion, the corresponding mass ablation rate is 0.452,0.455,0.432g/s and the line ablation rate is 1.863,1.323,0.843mm/s. The mass ablation rate is approximately unchanged and the line ablation rate is decreased. The smaller the ablation angle is, the more serious the thermalchemical ablation of gas, and thus increase the erosion area of axial fiber bundle under ablation condition and ablation gradient of radial fiber under ablation/erosion condition.
doi:10.11868/j.issn.1001-4381.2015.000527 doaj:864d3e4c02ae4d89b4c06d562a55711b fatcat:i6oldmnjabfwnb3fw3ybdvjoti

Riemannian Manifold Learning for Nonlinear Dimensionality Reduction [chapter]

Tony Lin, Hongbin Zha, Sang Uk Lee
2006 Lecture Notes in Computer Science  
In recent years, nonlinear dimensionality reduction (NLDR) techniques have attracted much attention in visual perception and many other areas of science. We propose an efficient algorithm called Riemannian manifold learning (RML). A Riemannian manifold can be constructed in the form of a simplicial complex, and thus its intrinsic dimension can be reliably estimated. Then the NLDR problem is solved by constructing Riemannian normal coordinates (RNC). Experimental results demonstrate that our
more » ... rithm can learn the data's intrinsic geometric structure, yielding uniformly distributed and well organized low-dimensional embedding data.
doi:10.1007/11744023_4 fatcat:tnopubylqzg37ex7gtd7o7dpcq

Essential Tensor Learning for Multi-view Spectral Clustering [article]

Jianlong Wu, Zhouchen Lin, Hongbin Zha
2019 arXiv   pre-print
Multi-view clustering attracts much attention recently, which aims to take advantage of multi-view information to improve the performance of clustering. However, most recent work mainly focus on self-representation based subspace clustering, which is of high computation complexity. In this paper, we focus on the Markov chain based spectral clustering method and propose a novel essential tensor learning method to explore the high order correlations for multi-view representation. We first
more » ... t a tensor based on multi-view transition probability matrices of the Markov chain. By incorporating the idea from robust principle component analysis, tensor singular value decomposition (t-SVD) based tensor nuclear norm is imposed to preserve the low-rank property of the essential tensor, which can well capture the principle information from multiple views. We also employ the tensor rotation operator for this task to better investigate the relationship among views as well as reduce the computation complexity. The proposed method can be efficiently optimized by the alternating direction method of multipliers~(ADMM). Extensive experiments on six real world datasets corresponding to five different applications show that our method achieves superior performance over other state-of-the-art methods.
arXiv:1807.03602v2 fatcat:nute6fz6t5fuvojg5trnlwhpxe

Theoretical Study of the Stability and Vibration Spectrum of C38Cl2

Zha Lin, Ma Guanhuai
2021 Science Discovery  
In this study, the theory of density functional was used to geometrically optimize the binary chlorine-derived isomers of C 38 at the b3lyp/6-31g(d) level. Analysis of the reaction heat and HOMO-LUMO of the optimized structure shows that C 38 Cl 2 is thermodynamically stable; The 1-8 added isomer 10-C 38 Cl 2 -2-37 and The 1-4 added isomer 17-C 38 Cl 2 -12-27 was the most stable among the molecules studied. Derivatization mostly occurs on the carbon atom with the most negative charge, which is
more » ... enerally the common vertex of three five membered rings. Chlorination derivatization reduces the HOMO energy level of the carbon cage, increases the HOMO-LUMO energy level gap of the carbon cage, and reduces the reaction activity of the carbon cage. The vibration frequency calculation shows that the addition mode and symmetry will affect the maximum vibration frequency and intensity of the structure. The strongest vibration is generally C-Cl telescopic vibration. The higher the derived position symmetry, the greater the vibration intensity.
doi:10.11648/ fatcat:o3zjqpummveqhek5aodldxmsjy

Industrial Efficiency Evaluation in China: A Nonparametric Production-Frontier Approach

Linlin Zhao, Lin Zhang, Yong Zha
2019 Sustainability  
An industrial system has positive and negative strategies to adapt to environmental regulations, which can be defined as natural disposability and managerial disposability. Meanwhile, the operational process of an industrial system can be divided into regular production activities and pollutant control activities. Within this, industrial system's technical efficiency (TE) can be decomposed into economic efficiency (ECE) and environmental efficiency (ENE). On the basis of natural disposability
more » ... d managerial disposability, this paper proposes static and dynamic data envelopment analysis (DEA) models to evaluate the efficiencies of industrial systems. Based on the proposed approach, TE, ECE, ENE, and Malmqusit productivity index (MPI) values were obtained simultaneously. The MPI values were further separated into the effects of static efficiency change and technical change. The proposed method was applied to assess the technical efficiencies of Chinese regional industrial systems between 2011 and 2015. Key findings are that (1) the low ENE is the main source of technical inefficiency; (2) the average static TE and ENE under natural disposability are both lower than those under managerial disposability; (3) the static efficiency change and technical change of TE are similar to those of ENE; and (4) the technical change has a significant impact on the changes in TE.
doi:10.3390/su11185019 fatcat:7kcrsdbogzhlziuodxk7fy2nle

CDP Mixture Models for Data Clustering

Yangfeng Ji, Tong Lin, Hongbin Zha
2010 2010 20th International Conference on Pattern Recognition  
In Dirichlet process (DP) mixture models, the number of components is implicitly determined by the sampling parameters of Dirichlet process. However, this kind of models usually produces lots of small mixture components when modeling real-world data, especially high-dimensional data. In this paper, we propose a new class of Dirichlet process mixture models with some constrained principles, named constrained Dirichlet process (CDP) mixture models. Based on general DP mixture models, we add a
more » ... mpling step to obtain latent parameters. In this way, CDP mixture models can suppress noise and generate the compact patterns of the data. Experimental results on data clustering show the remarkable performance of the CDP mixture models.
doi:10.1109/icpr.2010.161 dblp:conf/icpr/JiLZ10 fatcat:yy4ynrmw4nev3gzlo7g64gzkhy

VegaWarden: A Uniform User Management System for Cloud Applications

Jian Lin, Xiaoyi Lu, Lin Yu, Yongqiang Zou, Li Zha
2010 2010 IEEE Fifth International Conference on Networking, Architecture, and Storage  
In a virtual cluster based Cloud Computing environment, the sharing of infrastructure introduces two problems on user management: usability and security. Meanwhile, we observe that most conventional user management frameworks in the network environment are not fit for the scale expansion and interconnection of dynamic virtualization environment. In this paper, we propose VegaWarden, a uniform user management system to solve these problems. VegaWarden supplies a global user space for different
more » ... rtual infrastructures and application services in one Cloud, and allows user system interconnection among homogeneous Cloud instances. A uniform authentication model enables the security isolation of different administrative domains, and a decentralized architecture ensures its scalability. We have implemented VegaWarden in an experimental Cloud-oriented infrastructure and a production Grid Computing environment. The functionality and performance of VegaWarden have been demonstrated.
doi:10.1109/nas.2010.34 dblp:conf/nas/LinLYZZ10 fatcat:cjl5nmd3zrbx3mrf3axgxq7cqm
« Previous Showing results 1 — 15 out of 3,935 results