一个线性约束优化问题的广义梯度投影法

本文对线性约束优化问题提出了一个新的广义梯度投影法，该算法采用了非精确线性搜索，并在每次迭代运算中结合了广义投影矩阵和变尺度方法的思想确定其搜索方向．在通常的假设条件下，证明了该算法的整体收敛性和超线性收敛速度．     

解线性最小二乘问题的一个新并行算法

讨论了求解无约束线性最小二乘问题的一种并行单纯形法以及对它的改进算法并行共轭梯度—单纯形法 .算法本身具有很强的并行机制 ,能够充分地发挥并行机快速省时的特点 .本文也对算法做了理论分析 ,对算法的收敛性给予了证明 (在二维情形下 ) .最后做了数值实验 (由于软硬件条件的限制 ,并行算法未能在并行计算机上实现 ,鉴于这种情况 ,我们所做的数值实验均是在串行机上完成的 )     

Characterization of a 4-miktoarm star copolymer of the (PS-b-PI)3 PS type by temperature gradient interaction chromatography

Temperature gradient interaction chromatography (TGIC) was applied for the separation of a complex miktoarm star copolymer which has one polystyrene (PS) arm and three polystyrene-b-polyisoprene (PS-b-PI) diblock copolymer arms. Such miktoarm star polymers are much more difficult to characterize than branched homopolymers since the byproduct, typically polymers with missing arm(s) or coupled products, have not only different molecular weights but also different compositions. TGIC was able to fully separate the byproducts, and the composition of the molecular species corresponding to the different separated elution peaks was determined by two methods, fractionation/NMR and multiple detection (UV and RI). A reasonable agreement between the results of the two methods was obtained. By using the composition found, the corresponding molecular weights were determined by multi-angle light scattering detection. Based on the composition and the molecular weight we were able to identify the structure of the different molecular species.     

带惩罚项与随机输入的BP神经网络在线梯度学习算法的收敛性

本文对三层BP神经网络中带有惩罚项的在线梯度学习算法的收敛性问题进行了研究,在网络训练每一轮开始执行之前,对训练样本随机进行重排,以使网络学习更容易跳出局部极小,文中给出了误差函数的单调性定理以及该算法的弱收敛和强收敛性定理。     

基于MMD算法的HMM参数重估计的矩阵表示

本文提出的MMD算法用于提高模型区别错误信息和正确信息的能力．利用该算法在对模型的参数进行重估计时．涉及到复杂的目标函数的梯度运算．击运用矩阵运算使得梯度运算变得简单明了，因此本文给出了MMD算法下的HMM参数重估计的矩阵表示形式并给出了证明．     

线性均衡约束最优化的一个广义投影强次可行方向法

本文讨论带线性均衡约束最优化问题,首先利用摄动技术和一个互补函数将问题等价转化为一般约束最优化问题,然后结合广义投影技术和强次可行方向法思想,建立了问题的一个新算法.算法在迭代过程中保证搜索方向不为零,从而使得每次迭代只需计算一次广义投影.在适当的条件下,证明了算法的全局收敛性,并对算法进行了初步的数值试验.     

应用自动微分的Newton-PCG算法

一类新的使用符号微分的Newton-PCG型算法在文献[1]和[2]被导出来了。本文建立和研究应用自动微分的相应的Newton-PCG算法，理论分析和数值实验结果显示应用自动微分之后，目标函数的维数或复杂性越大，Newton-PCG算法对Newton法的改进越显著。     

二维椭圆型方程反问题中优化算法的比较

近年来，决定椭圆型方程系数反问题在地磁、地球物理、冶金和生物等实际问题上有着广泛的应用．本文讨论了二维的决定椭圆型方程系数反问题的数值求解方法．由误差平方和最小原则，这个反问题可化为一个变分问题，并进一步离散化为一个最优化问题，其目标函数依赖于要决定的方程系数．本文着重考察非线性共轭梯度法在此最优化问题数值计算中的表现，并与拟牛顿法作为对比．为了提高算法的效率我们适当选择加快收敛速度的预处理矩阵．同时还考察了线搜索方法的不同对优化算法的影响．数值实验的结果表明，非线性共轭梯度法在这类大规模优化问题中相对于拟牛顿法更有效．     

On the influence of the surfactant''s polar group on the local and terminal velocities of bubbles

The local and the terminal velocities, the size and the degree of bubbles’ shape deformations were determined as a function of distance from the position of the bubble formation (capillary orifice) in solutions of n-octyltrimethylammonium bromide, n-octyldimethylphosphine oxide, n-octyl-β-D-glucopyranoside and n-octanoic acid.

These surface-active compounds have different polar groups but an identical hydrocarbon chain (C8) in the molecule. The motion of the bubbles was monitored and recorded using a stroboscopic illumination, a CCD camera, and a JVC professional video. The recorded bubble images were analyzed by the image analysis software. The bubbles accelerated rapidly and their shape was deformed immediately after detachment from the capillary. The extent of the bubbles’ shape deformation (ratio of horizontal and vertical diameters) was 1.5 in distilled water and dropped rapidly down to a level of ca. 1.05–1.03 with increasing surfactant concentration. After the acceleration period the bubbles either attained a constant value of the terminal velocity (distilled water and high concentrations of the solutions), or a maximum in the velocity profiles was observed (low concentrations). The values of the terminal velocity diminished drastically with increasing concentration, from the value of 35 cm/s in water down to about 15 cm/s, while the bubble diameter decreased by ca. 10% only. The surfactant adsorption at the surface of the bubbles was evaluated and the minimum adsorption coverages required to immobilize the bubbles’ surface were determined. It was found that this minimum adsorption coverage was ca. 4% for n-octyldimethylphosphine oxide, n-octyl-β-D-glucopyranoside, n-octanoic acid and 25% for n-octyltrimethylammonium bromide. The difference in the adsorption coverage together with the surfactants’ surface activities indicate that it is mainly the adsorption kinetics of the surfactants that governs the fluidity of interfaces of the rising bubbles.     

Augmented Lagrangian Algorithms Based on the Spectral Projected Gradient Method for Solving Nonlinear Programming Problems

The spectral projected gradient method SPG is an algorithm for large-scale bound-constrained optimization introduced recently by Birgin, Martínez, and Raydan. It is based on the Raydan unconstrained generalization of the Barzilai-Borwein method for quadratics. The SPG algorithm turned out to be surprisingly effective for solving many large-scale minimization problems with box constraints. Therefore, it is natural to test its perfomance for solving the sub-problems that appear in nonlinear programming methods based on augmented Lagrangians. In this work, augmented Lagrangian methods which use SPG as the underlying convex-constraint solver are introduced (ALSPG) and the methods are tested in two sets of problems. First, a meaningful subset of large-scale nonlinearly constrained problems of the CUTE collection is solved and compared with the perfomance of LANCELOT. Second, a family of location problems in the minimax formulation is solved against the package FFSQP.