期刊文献+
共找到79篇文章
< 1 2 4 >
每页显示 20 50 100
TensorFlow平台下的手写字符识别 被引量:23
1
作者 张俊 李鑫 《电脑知识与技术》 2016年第6期199-201,共3页
基于谷歌第二代人工智能学习系统Tensor Flow,构建BP神经网络模型。将手写字符作为训练集输入神经网络,训练过程中不断调整权值和阈值,最终得到有较高识别精度的模型。体现了Tensor Flow在提高建模、编程、分析效率中的作用。通过此开... 基于谷歌第二代人工智能学习系统Tensor Flow,构建BP神经网络模型。将手写字符作为训练集输入神经网络,训练过程中不断调整权值和阈值,最终得到有较高识别精度的模型。体现了Tensor Flow在提高建模、编程、分析效率中的作用。通过此开发流程介绍,为进一步使用Tensor Flow构建复杂神经网络提供了参考。 展开更多
关键词 人工智能 tensor flow BP神经网络
下载PDF
Evolution of hypersurfaces by the mean curvature minus an external force field 被引量:11
2
作者 Yan-nan LIU Huai-yu JIAN 《Science China Mathematics》 SCIE 2007年第2期231-239,共9页
In this paper, we study the evolution of hypersurface moving by the mean curvature minus an external force field. It is shown that the flow will blow up in a finite time if the mean curvature of the initial surface is... In this paper, we study the evolution of hypersurface moving by the mean curvature minus an external force field. It is shown that the flow will blow up in a finite time if the mean curvature of the initial surface is larger than some constant depending on the boundness of derivatives of the external force field. For a linear force, we prove that the convexity of the hypersurface is preserved during the evolution and the flow has a unique smooth solution in any finite time and expands to infinity as the time tends to infinity if the initial curvature is smaller than the slope of the force. 展开更多
关键词 parabolic equation mean curvature flow maximum principle (for tensor) 35K45 53A05
原文传递
对使用TensorRT加速AI深度学习推断效率的探索 被引量:11
3
作者 施一飞 《科技视界》 2017年第31期26-27,共2页
本文对使用Tensor RT对目前流行的人工智能框架训练的神经网络模型的优化做了初步探索。目前基于人工智能的应用呈现爆炸式增长,图像识别、语音识别,自然语言处理、图像检索等功能已经成了人们日常必须的工具,随之而来对人工智能的推理... 本文对使用Tensor RT对目前流行的人工智能框架训练的神经网络模型的优化做了初步探索。目前基于人工智能的应用呈现爆炸式增长,图像识别、语音识别,自然语言处理、图像检索等功能已经成了人们日常必须的工具,随之而来对人工智能的推理效率和响应速度则提出更高要求。本文结合英伟达(n Vidia)的Tensor RT 3引擎和面向无人智能化的Jetson TX2平台,对以上问题的解决做了初步尝试。 展开更多
关键词 人工智能 tensor RT tensor flow 神经网络
下载PDF
基于深度学习的高分遥感影像乡镇建筑物识别方法 被引量:7
4
作者 王利忠 张宏海 +1 位作者 仲波 牛铁 《科研信息化技术与应用》 2019年第1期88-95,共8页
目的随着遥感卫星技术的进步,卫星图像的空间分辨率、时间分辨率及光谱分辨率越来越高,所包含的信息越来越丰富。乡镇建筑物的识别可以利用高分辨率遥感影像进行提取,可以有效地分析农村的人口密度和用地分配,从而为农村发展规划提供重... 目的随着遥感卫星技术的进步,卫星图像的空间分辨率、时间分辨率及光谱分辨率越来越高,所包含的信息越来越丰富。乡镇建筑物的识别可以利用高分辨率遥感影像进行提取,可以有效地分析农村的人口密度和用地分配,从而为农村发展规划提供重要参考。方法根据乡镇建筑物的特点将建筑物分成对应的几类进行标注,然后将样本送入已经搭建好的基于Tensor Flow平台的FasterRCNN模型进行训练,利用CNN提取特征,通过构建区域推荐网络(RPN)提取可能的建筑物区域,利用检测网络对目标区域进行判别和分类。结果最终模型的识别准确率达到93.06%,具有较好的鲁棒性和泛化性。结论本文提出的乡镇建筑物提取框架,建立了乡镇区域的建筑物样本库,有效地避免了图像分类的局限性,同时提高了乡镇建筑物识别的精度。 展开更多
关键词 卫星图像 乡镇建筑物 Faster-RCNN 区域推荐网络 tensor flow
原文传递
基于TensorFlow的LSTM循环神经网络短期电力负荷预测 被引量:7
5
作者 李松岭 《上海节能》 2018年第12期974-977,共4页
随着智能电网、坚强电网的建立及人工智能领域技术的高速发展,如何对电力领域的负荷进行更高精度的预测已成为电力从业者们特别关注与研究的问题。基于TensorFlow智能学习系统的深度学习LSTM循环神经网络算法的短期电力负荷预测算法,结... 随着智能电网、坚强电网的建立及人工智能领域技术的高速发展,如何对电力领域的负荷进行更高精度的预测已成为电力从业者们特别关注与研究的问题。基于TensorFlow智能学习系统的深度学习LSTM循环神经网络算法的短期电力负荷预测算法,结合某地区发电厂负荷数据设计实验,通过多次数据迭代、参数更新,进行模型训练与预测,最终的实验证明:基于TensorFlow的LSTM循环神经网络算法预测效果明显好于传统机器学习算法。随着数据量的增大,模型更显示出其良好的鲁棒性。 展开更多
关键词 tensorflow LSTM 深度学习 短期电力负荷预测
下载PDF
采动覆岩微震分区演化特征的数值模拟研究 被引量:6
6
作者 袁国涛 张明伟 +2 位作者 王杰 卫俊 杨坤 《煤炭科学技术》 EI CAS CSCD 北大核心 2023年第8期36-46,共11页
为了探究采动覆岩的破裂分区特征与微震演化规律,以某矿井工程地质条件为背景,根据矩张量与颗粒流理论构建了微震模拟方法,实现了对回采工作面覆岩微震演化特征的模拟研究。研究表明:采动覆岩的微震事件矩震级集中在–2.7~–1,微震破裂... 为了探究采动覆岩的破裂分区特征与微震演化规律,以某矿井工程地质条件为背景,根据矩张量与颗粒流理论构建了微震模拟方法,实现了对回采工作面覆岩微震演化特征的模拟研究。研究表明:采动覆岩的微震事件矩震级集中在–2.7~–1,微震破裂强度符合常见的正态分布性质,微震发生频次与微震矩震级的关系也满足经典的矩震级–频度关系式;微震事件在采场横向上表现出明显的区域分布特征,采空区左右煤壁中的微震事件先于采空区顶板发生,微震事件围绕采掘空间向两侧发展,且存在一定的超前性与滞后性;当工作面开采一定距离后,采空区上覆岩层的垂向应力表现出“双峰值”特征,微震事件迅速向上延伸,并在上覆软弱岩层内集聚、发育,使得微震事件在竖向上同样表现出明显的分层分布特征;根据采场的应力、位移以及微震分布特征,可以将上覆岩层划分为“横四区”与“竖三带”,各个破裂区域的微震事件破裂类型占比各不相同,煤壁支撑区与离层区分别为剪切型与拉伸型微震事件的主要占比区域。此外,对研究区域建立了微震监测系统,现场监测与数值模拟结果基本吻合,表明构建的微震模拟方法适用性较好,可以从细观层面再现采动覆岩的微震演化过程,研究结果为揭示采动裂隙演化机理提供一定的理论基础。 展开更多
关键词 采动覆岩 微震 矩张量 颗粒流 破裂类型
下载PDF
基于TensorFlow的交通标志识别方法研究 被引量:5
7
作者 王全 梁敬文 《价值工程》 2019年第27期204-206,共3页
交通标志识别系统是智能驾驶系统的重要组成部分;本文分析了现有方法存在的问题,基于TensorFlow框架搭建了改进的卷积神经网络,用于识别交通标志;整个系统在TensorFlow上实现,使用行车记录仪采集的视频验证了本文的算法,结果表明本文算... 交通标志识别系统是智能驾驶系统的重要组成部分;本文分析了现有方法存在的问题,基于TensorFlow框架搭建了改进的卷积神经网络,用于识别交通标志;整个系统在TensorFlow上实现,使用行车记录仪采集的视频验证了本文的算法,结果表明本文算法有一定的实用性,而且在准确率,鲁棒性和实时性等方面也表现较好。 展开更多
关键词 交通标志识别 卷积神经网络 tensor flow
下载PDF
NUMERICAL SIMULATION OF MORPHOLOGY OF POLYMER CHAIN COILS IN COMPLEX FLOWS 被引量:4
8
作者 Hong Zheng Wei Yu Chi-xing Zhou Department of Polymer Materials Science & Engineering,Shanghai Jiaotong University,Shanghai 200240,China 《Chinese Journal of Polymer Science》 SCIE CAS CSCD 2005年第5期453-462,共10页
A new coupled finite element formulation is proposed to calculate a conformation tensor model in two complex flows: a planar contraction flow and a planar flow around a symmetrically placed cylinder. The components o... A new coupled finite element formulation is proposed to calculate a conformation tensor model in two complex flows: a planar contraction flow and a planar flow around a symmetrically placed cylinder. The components of conformation tensor are first computed together with the velocity and pressure to describe the change of morphology of polymer chain coils in flow fields. Macroscopic quantities of viscoelastic flow are then calculated based on the conformation tensor. Comparisons between the numerical simulations and experiments for stress patterns and velocity profiles are carried out to prove the validity of the method. 展开更多
关键词 Viscoelastic flow MORPHOLOGY Complex flows Conformation tensor model Finite element method (FEM).
下载PDF
An Efficient Numerical Model for Immiscible Two-Phase Flow in Fractured Karst Reservoirs 被引量:4
9
作者 Zhao-Qin Huang Jun Yao Yue-Ying Wang 《Communications in Computational Physics》 SCIE 2013年第2期540-558,共19页
Numerical simulation of two-phase flow in fractured karst reservoirs is still a challenging issue.The triple-porosity model is the major approach up to now.However,the triple-continuum assumption in this model is unac... Numerical simulation of two-phase flow in fractured karst reservoirs is still a challenging issue.The triple-porosity model is the major approach up to now.However,the triple-continuum assumption in this model is unacceptable for many cases.In the present work,an efficient numerical model has been developed for immiscible two-phase flowin fractured karst reservoirs based on the idea of equivalent continuum representation.First,based on the discrete fracture-vug model and homogenization theory,the effective absolute permeability tensors for each grid blocks are calculated.And then an analytical procedure to obtain a pseudo relative permeability curves for a grid block containing fractures and cavities has been successfully implemented.Next,a full-tensor simulator has been designed based on a hybrid numerical method(combining mixed finite element method and finite volume method).A simple fracture system has been used to demonstrate the validity of our method.At last,we have used the fracture and cavity statistics data fromTAHE outcrops in west China,effective permeability values and other parameters from our code,and an equivalent continuum simulator to calculate the water flooding profiles for more realistic systems. 展开更多
关键词 Fractured karst reservoirs effective permeability tensor discrete fracture-vug network model two-phase flow full-tensor simulator
原文传递
ON THE KHLER-RICCI SOLITONS WITH VANISHING BOCHNER-WEYL TENSOR 被引量:3
10
作者 苏延辉 张坤 《Acta Mathematica Scientia》 SCIE CSCD 2012年第3期1239-1244,共6页
In this article, we study the steady, shrinking, and expanding Kahler-Ricci solitons with vanishing Bochner-Weyl tensor and prove that, under this condition, the Ricci solitons must have constant holomorphic sectional... In this article, we study the steady, shrinking, and expanding Kahler-Ricci solitons with vanishing Bochner-Weyl tensor and prove that, under this condition, the Ricci solitons must have constant holomorphic sectional curvature. 展开更多
关键词 Ricci flow Kahler Ricci soliton Bochner-Weyl tensor
下载PDF
基于TensorFlow的手写数字识别系统 被引量:4
11
作者 陈昊 郭海 +1 位作者 刘大全 张家旗 《信息通信》 2018年第3期108-110,共3页
为制作脱机手写数字识别系统,运用机器学习理论,在Tensor Flow平台下对K最近邻挖掘、循环神经网络、卷积神经网络等诸多算法进行了比对测试,最后以测试精度达到99.36%的卷积神经网络为核心算法制作了操作简便的手写数字识别系统。
关键词 手写识别 tensor flow 卷积神经网络
下载PDF
一种基于TensorFlow的DCGAN模型实现 被引量:4
12
作者 樊雷 《电脑知识与技术》 2017年第12X期219-221,共3页
卷积神经网络CNN目前在有监督学习领域有着优秀的表现,但在无监督学习领域研究进展缓慢。该文将CNN引入到GAN中的生成式模型中做无监督训练,利用CNN强大的特征提取能力来提高生成式模型的学习效果,采用TensorFlow和Python代码实现了DCGA... 卷积神经网络CNN目前在有监督学习领域有着优秀的表现,但在无监督学习领域研究进展缓慢。该文将CNN引入到GAN中的生成式模型中做无监督训练,利用CNN强大的特征提取能力来提高生成式模型的学习效果,采用TensorFlow和Python代码实现了DCGAN中的D模型和G模型,并在MNIST部分数据集下验证了模型生成数字图像效果。实验结果表明采用DCGAN可以有效获取图像表征用于分类并具备生成较高分辨率的图像能力。 展开更多
关键词 DCGAN 深度学习 tensorflow GAN
下载PDF
采用GPS观测值进行三维形变分析的构想 被引量:3
13
作者 刘序俨 梁全强 +1 位作者 黄声明 丁学仁 《大地测量与地球动力学》 CSCD 北大核心 2006年第1期92-96,共5页
在已知地表度量张量的基础上,给出了根据测站的GPS复测值计算地表形变张量、主方向、线应变、面应变、高斯曲率与平均曲率的有关公式,并着重对地表的内蕴与外蕴形变概念作了阐述,最后指出利用地表内蕴与外蕴形变量能较好地描述地表的三... 在已知地表度量张量的基础上,给出了根据测站的GPS复测值计算地表形变张量、主方向、线应变、面应变、高斯曲率与平均曲率的有关公式,并着重对地表的内蕴与外蕴形变概念作了阐述,最后指出利用地表内蕴与外蕴形变量能较好地描述地表的三维形变。 展开更多
关键词 GPS观测值 度量张量 内蕴与外蕴形变 三维形变 分析流程
下载PDF
The Research of the Flow of Bodies with Use of the Vector Form of the Newton’s Law for the Viscous Liquid 被引量:2
14
作者 Andrey N. Volobuev Eugene S. Petrov 《Engineering(科研)》 2011年第2期162-167,共6页
On the basis of the vector formula of the Newton’s law for a viscous liquid and the integrated vector form of the equation of an impulse for a viscous liquid for resistance and carrying power of a profile of any form... On the basis of the vector formula of the Newton’s law for a viscous liquid and the integrated vector form of the equation of an impulse for a viscous liquid for resistance and carrying power of a profile of any form and the big length dependences are found in a stream. Application of the found dependences at a circulating flow of the cylinder located across a stream is showed. The analysis of a tensor of viscosity for laminar and turbulent flow is carried out. 展开更多
关键词 Newton’s LAW for VISCOSITY A Body flow CIRCULATION of Speed Resistance Factor A VISCOSITY tensor
下载PDF
基于多任务级联卷积网络模型的人脸检测和识别 被引量:2
15
作者 刘其嘉 郭一娜 +1 位作者 任晓文 李健宇 《太原科技大学学报》 2019年第2期81-85,共5页
深度学习的集成特征提取这一优点使得它广泛应用于人脸检测和识别。提出了一种多任务级联卷积网络模型(Multitask Cascaded Convolution Network,MTCNN)。基于Tensor Flow平台,基于改进的任务级联卷积网络模型检测到人脸,并且用Face Ne... 深度学习的集成特征提取这一优点使得它广泛应用于人脸检测和识别。提出了一种多任务级联卷积网络模型(Multitask Cascaded Convolution Network,MTCNN)。基于Tensor Flow平台,基于改进的任务级联卷积网络模型检测到人脸,并且用Face Net算法对人脸进行特征提取,用KNN算法对人脸进行识别。实验结果表明,对不同光照下多人图像和遮挡图像的人脸进行检测和识别,具有良好的鲁棒性。 展开更多
关键词 人脸检测和识别 深度学习 tensor flow 多任务级联卷积网络 FACE Net
下载PDF
Complete bounded λ-hypersurfaces in the weighted volume-preserving mean curvature flow 被引量:3
16
作者 Yecheng Zhu Yi Fang Qing Chen 《Science China Mathematics》 SCIE CSCD 2018年第5期929-942,共14页
In this paper, we study the complete bounded λ-hypersurfaces in the weighted volume-preserving mean curvature flow. Firstly, we investigate the volume comparison theorem of complete bounded λ-hypersurfaces with |A|... In this paper, we study the complete bounded λ-hypersurfaces in the weighted volume-preserving mean curvature flow. Firstly, we investigate the volume comparison theorem of complete bounded λ-hypersurfaces with |A|≤α and get some applications of the volume comparison theorem. Secondly, we consider the relation among λ, extrinsic radius k, intrinsic diameter d, and dimension n of the complete λ-hypersurface,and we obtain some estimates for the intrinsic diameter and the extrinsic radius. At last, we get some topological properties of the bounded λ-hypersurface with some natural and general restrictions. 展开更多
关键词 volume comparison theorem topology second fundamental form ∞-Bakry-Emery Ricci tensor mean curvature flow
原文传递
卷积神经网络图像识别技术在入侵监测系统中的应用研究 被引量:3
17
作者 王洪德 王焕青 《大连交通大学学报》 CAS 2017年第5期107-110,共4页
为应对入侵手段复杂多样的安全形势,解决现有入侵监测技术成本高、适应性差的问题,设计研发一种成本低廉、兼容性好、方便拓展的入侵监测系统.应用开源深度机器学习框架Tensor FLow实现基于卷积神经网络的图像识别算法,构建辅助安全系统... 为应对入侵手段复杂多样的安全形势,解决现有入侵监测技术成本高、适应性差的问题,设计研发一种成本低廉、兼容性好、方便拓展的入侵监测系统.应用开源深度机器学习框架Tensor FLow实现基于卷积神经网络的图像识别算法,构建辅助安全系统,并在多种工况下对系统有效性进行验证.结果表明:系统能以较高准确率对入侵行为进行识别,在多分类情景下,随训练样本数量的增加,模型预测准确率得到提高,而收敛时间有所增加,开启GPU加速后缩短为原来的1/10. 展开更多
关键词 入侵监测 图像识别 卷积神经网络 tensor flow
下载PDF
OVERSHOOTS IN STRESS AND FREE ENERGY CHANGE DURING THE FLOWINDUCED CRYSTALLIZATION OF POLYMERIC MELT IN SHEAR FLOW 被引量:1
18
作者 张洪斌 《Chinese Journal of Polymer Science》 SCIE CAS CSCD 2010年第4期657-666,共10页
The effect of pre-shear flow on the subsequent crystallization process of polymeric melt was investigated and a flow-induced crystallization (FIC) model based on the conformation tensor incorporating the pre-shear e... The effect of pre-shear flow on the subsequent crystallization process of polymeric melt was investigated and a flow-induced crystallization (FIC) model based on the conformation tensor incorporating the pre-shear effect was proposed. The model is capable of predicting the overshoot phenomena of the stress and the flow-induced free energy change of the polymeric system at high pre-shear rates. Under the condition of flow, the increase in the activated nuclei number was contributed by the flow-induced free energy change, which showed an overwhelming effect on the nuclei formation during the pre-shear process at high shear rates. The half crystallization time (t1/2) of polypropylene (PP) as functions of pre-shear rate and pre-shear time at different crystallization temperatures was predicted and compared with the experiment data. Both numerical and experimental results showed that t1/2 of PP decreased dramatically when the flow started but leveled off at long times. It was found that two transformation stages in t1/2 existed within a wide range of shear rates. For the first stage where the melting polymer experienced a relatively weak shear flow, the acceleration of crystallization kinetics was mainly contributed by the steady value of free energy change while in the second stage for high shear rates, strong overshoot in flow-induced free energy change occurred and the crystallization kinetics was thus significantly enhanced. The overshoots in stress and flow-induced free energy change reflected an important role of flow on the primary nucleation especially when the flow was strong enough. 展开更多
关键词 Crystallization kinetics Half crystallization time flow Model Conformation tensor
原文传递
New conception in continuum theory of constitutive equation for anisotropic crystalline polymer liquids 被引量:1
19
作者 Shifang Han 《Natural Science》 2010年第9期948-958,共11页
A new continuum theory of the constitutive equation of co-rotational derivative type is developed for anisotropic viscoelastic fluid—liquid crystalline (LC) polymers. A new concept of simple anisotropic fluid is intr... A new continuum theory of the constitutive equation of co-rotational derivative type is developed for anisotropic viscoelastic fluid—liquid crystalline (LC) polymers. A new concept of simple anisotropic fluid is introduced. On the basis of principles of anisotropic simple fluid, stress behaviour is described by velocity gradient tensor and spin tensor instead of the velocity gradient tensor in the classic Leslie—Ericksen continuum theory. Analyzing rheological nature of the fluid and using tensor analysis a general form of the constitutive equ- ation of co-rotational type is established for the fluid. A special term of high order in the equation is introduced by author to describe the sp- ecial change of the normal stress differences which is considered as a result of director tumbling by Larson et al. Analyzing the experimental results by Larson et al., a principle of Non- oscillatory normal stress is introduced which leads to simplification of the problem with relaxation times. The special behaviour of non- symmetry of the shear stress is predicted by using the present model for LC polymer liquids. Two shear stresses in shear flow of LC polymer liquids may lead to vortex and rotation flow, i.e. director tumbling in the flow. The first and second normal stress differences are calculated by the model special behaviour of which is in agree- ment with experiments. In the research, the com- putational symbolic manipulation such as computer software Maple is used. For the anisotropic viscoelastic fluid the constitutive equation theory is of important fundamental significance. 展开更多
关键词 Constitutive Equation Co-rotational Derivative Type Simple ANISOTROPIC FLUID Non-Newtonian FLUID Liquid Crystalline Polymer NORMAL STRESS Difference SHEAR flow Un-symmetry of SHEAR STRESS tensor Director TUMBLING Effect Non-Oscillatory NORMAL STRESS
下载PDF
IMPLEMENTATION OF MIXED METHODS AS FINITE DIFFERENCE METHODS AND APPLICATIONS TO NONISOTHERMAL MULTIPHASE FLOW IN POROUS MEDIA 被引量:1
20
作者 Zhang-xin Chen Xi-jun Yu 《Journal of Computational Mathematics》 SCIE CSCD 2006年第3期281-294,共14页
In this paper we consider mixed finite element methods for second order elliptic problems. In the case of the lowest order Brezzi-Douglas-Marini elements (if d = 2) or Brezzi- Douglas-Duran-Fortin elements (if d = ... In this paper we consider mixed finite element methods for second order elliptic problems. In the case of the lowest order Brezzi-Douglas-Marini elements (if d = 2) or Brezzi- Douglas-Duran-Fortin elements (if d = 3) on rectangular parallelepipeds, we show that the mixed method system, by incorporating certain quadrature rules, can be written as a simple, cell-centered finite difference method. This leads to the solution of a sparse, positive semidefinite linear system for the scalar unknown. For a diagonal tensor coefficient, the sparsity pattern for the scalar unknown is a five point stencil if d = 2, and seven if d = 3. For a general tensor coefficient, it is a nine point stencil, and nineteen, respectively. Applications of the mixed method implementation as finite differences to nonisothermal multiphase, multicomponent flow in porous media are presented. 展开更多
关键词 Finite difference IMPLEMENTATION Mixed method Error estimates SUPERCONVERGENCE tensor coefficient Nonisothermal multiphase Multicomponent flow Porousmedia.
原文传递
上一页 1 2 4 下一页 到第
使用帮助 返回顶部