采用几何模型切分的人脸纹理图像生成算法

为了克服传统的照片图像拼接方法中利用特征线进行不同照片之间公共交界线定位不准确的缺点,提出一种"几何模型切分"的人脸纹理图像生成算法。通过对人脸几何模型进行切分,以切分后模型图片的轮廓作为边界线裁剪相应的人脸照片,实现不同照片之间交界线的准确对接,并采用柱面纹理映射方法将生成的纹理图像映射到特定人脸几何模型上。实验结果表明,采用提出的"几何模型切分"算法生成的人脸纹理图像进行纹理映射可以得到较好的真实感三维人脸模型,是一种生成人脸纹理图像的有效方法。

深度神经网络模型任务切分及并行优化方法

为解决传统手工切分神经网络模型计算任务并行化方法面临的并行化难度大、训练耗时长、设备利用率低等问题,提出了一种基于深度神经网络(DNN)模型特性感知的任务切分及并行优化方法。结合硬件计算环境,对模型计算特性进行动态分析,获取模型内部相关性和各类参数属性,构建原始计算任务有向无环图(DAG);利用增强反链,构建DAG节点间可分区聚类的拓扑关系,将原始DAG转换为易于切分的反链DAG;通过拓扑排序生成反链DAG状态序列,并使用动态规划将状态序列切分为不同执行阶段,分析最佳分割点进行模型切分,实现模型分区与各GPU间动态匹配;对批量进行微处理,通过引入流水线并行实现多迭代密集训练,提高GPU利用率,减少训练耗时。实验结果表明:与已有模型切分方法相比,在CIFAR-10数据集上,所提模型切分及并行优化方法可实现各GPU间训练任务负载均衡,在保证模型训练精度的同时,4 GPU加速比达到3.4,8 GPU加速比为3.76。

基于光刻模型的光学邻近校正切分优化方法

在处理二维图形时需要耗费大量时间调试配方。为此,提出一种基于光刻模型的切分优化方法。采用纹波抑制的切分方法,并根据空间光强曲线的极值点分配切分片段,通过格点搜索选取方法参数,从而获得最优切分方案。实验结果表明,该方法能获得稳定的配方调试时间,且与参考配方结果相比,具有较好的光学邻近校正精度,可使边位置误差平均降低5%。

层次短语翻译中基于Markov随机场的层次切分模型

翻译推导的切分歧义是统计机器翻译面临的一个很重要的问题,而在层次短语机器翻译中,其尤为突出.提出了一个层次切分模型来处理推导的切分歧义性.采用Markov随机场构建模型,然后将其融入层次短语翻译模型,以便自动选择更合理的切分.在NIST中英翻译的任务中,该模型的训练效率高,通过NIST05,NIST06和NIST08这3个测试集上的翻译效果表明,该模型提高了层次短语翻译的性能.

改进机器翻译中的句子切分模型

随着统计机器翻译系统训练语料的不断增加,长句子的数量越来越多,如何有效地利用长句子中的信息改进翻译质量是统计机器翻译系统面临的主要问题之一。该文基于Xu的句子切分模型,提出了一种在训练阶段切分长句子的方法,该方法利用自动获取的边界词概率和切分后子句对的长度比例来指导切分过程,从而得到更符合语义信息的句子切分结果。在NIST测试集上的实验结果表明,该方法获得了最大0.5个BLEU值的提升。

基于贝叶斯方法的鲁棒语音切分

在基于隐马尔科夫模型的语音切分基础上 ,融合了不受噪声干扰的先验切分模型 ,提出了基于贝叶斯方法的语音切分方法。在贝叶斯切分方法的框架内 ,作者首先对语音序列进行了变换 ,将由切分点构成的序列变为由音节长度构成的序列。然后 ,假设音节长度序列符合一阶马尔科夫过程 ,经过归一化处理后 ,求出了切分的先验概率公式 ,得到了贝叶斯方法的切分模型。在噪声环境下的实验证明 ,由于切分模型独立于噪声 ,对在噪声环境下声学模型的失配提供了很好的补偿 ,使得语音切分的鲁棒性大大增加。

Feature study for improving Chinese overlapping ambiguity resolution based on SVM

In order to improve Chinese overlapping ambiguity resolution based on a support vector machine, statistical features are studied for representing the feature vectors. First, four statistical parameters-mutual information, accessor variety, two-character word frequency and single-character word frequency are used to describe the feature vectors respectively. Then other parameters are tried to add as complementary features to the parameters which obtain the best results for further improving the classification performance. Experimental results show that features represented by mutual information, single-character word frequency and accessor variety can obtain an optimum result of 94. 39%. Compared with a commonly used word probability model, the accuracy has been improved by 6. 62%. Such comparative results confirm that the classification performance can be improved by feature selection and representation.

Analytical model of shear mechanical behaviour of bolted rock joints considering influence of normal stress on bolt guide rail effect

Rock bolts have been widely used in slopes as a reinforcement measure.Modelling the shear mechanical behaviours of bolted rock joints is very complicated due to the complex factors that affect the axial force and shear force on the bolts.Rock bolts under shear action exhibit the guide rail effect;that is,the rock mass slides along the rock bolt as if the rock bolt is a rail.The normal stress can inhibit the guide rail effect and reduce the axial force on bolts.However,this factor is not considered by the existing analysis models.Shear tests of bolted joints under different normal stresses were carried out in the laboratory.During the test,the axial force on each point monitored on the bolt was recorded by a strain gauge,and the attenuation trend of the strain was studied.An analytical model that considers the inhibition of the bolt rail effect due to an increase in the normal stress was proposed to predict the shear mechanical behaviour of rock bolted joints.The new model accommodates the bolt shear behaviours in the elastic stage and plastic stage,and the estimated values agree well with the results of the direct shear tests in the laboratory.The validation shows that the proposed model can effectively describe the deformation characteristics of the bolts in the shear tests.

A lexicalized second-order-HMM for ambiguity resolution in Chinese segmentation and POS tagging

Hidden Markov Model（HMM） is a main solution to ambiguities in Chinese segmentation anti POS （part-of-speech） tagging. While most previous works tot HMM-based Chinese segmentation anti POS tagging eonsult POS informatiou in contexts, they do not utilize lexieal information which is crucial for resoMng certain morphologieal ambiguity. This paper proposes a method which incorporates lexieal information and wider context information into HMM. Model induction anti related smoothing technique are presented in detail. Experiments indicate that this technique improves the segmentation and tagging accuracy by nearly 1%.

Development of a one-axis manipulator for laser-assisted machining

Laser-assisted machining has been considered as a new alternative machining method of difficult-to-cut materials. A laser module with one-axis manipulator is developed to focus on preheating laser beam effectively. First of all, the thermal characteristic analysis was performed to verify the importance of laser module location. Laser module should be moved within 1 mm. Analysis conditions of three positions in driving range of the one-axis manipulator are selected. And a C coupling is used as a connection device for spindle and laser module. An initial model has one C coupling, and the number of C coupling has been increased from 1 to 2 in an improved model. And the analysis is carried out again for the one-axis manipulator. The results of the static analysis, the maximum displacement and the maximum stress are decreased by 22% and 11%, respectively, for the improved model when the laser module is located at farthest away from the spindle unit. As a result of the modal analysis, the first natural frequency mode is increased by 13%, 18% and 12% at these positions of the improved model, respectively. The harmonic analysis of the improved model was performed by analyzing the results of the modal analysis. The maximum deformation was 0.33 mm in driving unit at 222 Hz. And the maximum compliance of the ISO axis direction was 0.23 mm/N. Finally, the one-axis manipulator has been fabricated successfully using the analysis result.

Cutting edge curve models for equal pitch cutters and their applications

A mathematic model is established using infinitesimal geometry for the cutting edge design of special milling cutters which use equal lead helix as cutting edges; equations are given for front-end and proclitic surface of revolution of ball pillar milling cutters, ball taper milling cutters and angularly conical milling cutters; and corresponding models are established for the continuity cutting edge curves of milling cutters. Typical examples are given to illustrate the applications of mathematic models, which prove the correctness and applicability of these geometric models.

ARX-NNPLS Model Based Optimization Strategy and Its Application in Polymer Grade Transition Process

Since it is often difficult to build differential algebraic equations (DAEs) for chemical processes, a new data-based modeling approach is proposed using ARX (AutoRegressive with eXogenous inputs) combined with neural network under partial least squares framework (ARX-NNPLS), in which less specific knowledge of the process is required but the input and output data. To represent the dynamic and nonlinear behavior of the process, the ARX combined with neural network is used in the partial least squares (PLS) inner model between input and output latent variables. In the proposed dynamic optimization strategy based on the ARX-NNPLS model, neither parameterization nor iterative solving process for DAEs is needed as the ARX-NNPLS model gives a proper representation for the dynamic behavior of the process, and the computing time is greatly reduced compared to conventional control vector parameterization method. To demonstrate the ARX-NNPLS model based optimization strategy, the polyethylene grade transition in gas phase fluidized-bed reactor is taken into account. The optimization results show that the final optimal trajectory of quality index determined by the new approach moves faster to the target values and the computing time is much less.

Reliability analysis of laminated composite under compression and shear loads

Carrying on a series of compression and shear tests by a large number of specimens, reliabilities of T300/QY8911 laminated composite were studied when dispersibility models were described. The results show that the stress is linearly dependent on the strain and the damage modes of specimens are brittle fracture for both kinds of tests. Dispersibility models of compression and shear strength are expressed as Re-N（415.39, 6 586.36） and Rs-ln（5.071 8, 0.155 3）, respectively. When normal and lognormal distributions were used to describe the dispersibility models of compression and shear strength, and the compression or shear load follows the normal distribution, the almost same failure probability can be obtained from different reliability analysis methods.

Evaluation and Optimization of the Mixed Redundancy Strategy in Cloud-Based Systems

Mixed redundancy strategies are generally used in cloud-based systems,with different node switch mechanisms from traditional fault-tolerant strategies.Existing studies often concentrate on optimizing a single strategy in cloud computing environment and ignore the impact of mixed redundancy strategies.Therefore,a model is proposed to evaluate and optimize the reliability and performance of cloud-based degraded systems subject to a mixed active and cold standby redundancy strategy.In this strategy,node switching is triggered by a continual monitoring and detection mechanism when active nodes fail.To evaluate the transient availability and the expected job completion rate of systems with such kind of strategy,a continuous-time Markov chain model is built on the state transition process and a numerical method is used to solve the model.To choose the optimal redundancy for the mixed strategy under system constraints,a greedy search algorithm is proposed after sensitivity analysis.Illustrative examples were presented to explain the process of calculating the transient probability of each system state and in turn,the availability and performance of the whole system.It was shown that the near-optimal redundancy solution could be obtained using the optimizationmethod.The comparison with optimization of the traditional mixed redundancy strategy proved that the system behavior was different using different kinds of mixed strategies and less redundancy was assigned for the new type of mixed strategy under the same system constraint.

Computational Fluid Dynamics Simulation on Biomedical Stent Design

The stent was a major breakthrough in the treatment of atherosclerotic vascular disease. The permanent vascular implant of a stent, however, changes the intra-stent blood flow hemodynamics. There is a growing consensus that the stent implant may change the artery wall shear stress distribution and hence lead to the restenosis process. Computational fluid dynamics （CFD） has been widely used to analyze hemodynamics in stented arteries. In this paper, two CFD models （the axisymmetric model and the 3-D stent model） were developed to investigate the effects of strut geometry and blood rheology on the intra-stent hemodynamics. The velocity profile, flow recirculation, and wall shear stress distribution of various stent strut geometries were studied. Results show strong correlations between the intra-stent hemodynamics and strut geometry. The intra-stent blood flow is very sensitive to the strut height and fillet size. A round strut with a large fillet size shows 36% and 34% reductions in key parameters evaluating the restenosis risk for the axisymmetric model and the 3-D stent model, respectively. This suggests that electrochemical polishing, a surface-improving process during stent manufacturing, strongly influences the hemodynamic behavior in stented arteries and should be controlled precisely in order to achieve the best clinical outcome. Rheological effects on the wall shear stress are minor in both axisymmetric and 3-D stent models for the vessel diameter of 4 mm, with Newtonian flow simulation tending to give more conservative estimates ofrestenosis risk. Therefore, it is reasonable to simulate the blood flow as a Newtonian flow in stented arteries using the simpler axisymmetric model. These findings will provide great insights for stent design optimization for potential restenosis improvement.

Exact Solitary Wave Solutions of Nonlinear Evolution Equations with a Positive Fractional Power Term

The bounded and smooth solitary wave solutions of 10 nonlinear evolution equations with a positive fractional power term of dependent variable are successfully obtained by homogeneous balance principle and with the aid of sub-ODEs that admits a solution of sech-power or tanh-power type.In the special cases that the fractional power equals to 1 and 2,the solitary wave solutions of more than 10 important model equations arisen from mathematical physics are easily rediscovered.