本文研究小参数椭圆-抛物偏微分方程一致收敛差分格式的充要条件。
This paper studies the necessary and sufficient condition of uniformly convergent difference scheme for the elliptic-parabolic partial differential equation with a small parameter.
在基本差分进化算法中,融入递增二次函数交叉算子以增加算法的收敛速度。
For the basic differential evolution algorithm, the increasing quadratic function crossover operator was added to increase the convergence speed.
当在空间位置出现突变时,偏微分方程的经典解法如有限元法和加权残差法等无法保证解的精度和收敛性。
The problems of resolution and convergence are increased by classic methods such as FEM and WRM when the solutions exhibit sharp changes.
证明了差分格式的收敛性与稳定性。
数值结果显示了该方法的优越性,包括迭代序列的单调收敛性及有限差分解的高精度。
The numerical results demonstrate the advantages of the method, including the monotone convergence property of iterative sequences and the high accuracy of the method.
我们利用问题的渐近解证明了差分格式关于小参数的一致收敛性。
By means of the asymptotic solution of singular perturbation problem we proved the uniform convergence of this scheme with respect to the small parameter.
该算法克服了DFS算法收敛性差和模拟退火(SA)算法收敛速度慢的缺点。
Compared with traditional DFS algorithm and simulated annealing (SA) algorithm, the proposed algorithm possesses better convergency and high convergence rate.
还指出由于当前的连续小波神经网络主要使用传统BP神经网络的随机初始化方法和基于梯度的训练算法,因此存在收敛性差的缺点。
It is also indicated that current WNN has a poor convergence performance because of adopting the random initialization method and gradient training algorithm of traditional BP NET.
介绍一种具有“收敛”特性的有限区域旋转风与辐散风分解方案,并与目前常用的一些差分方案进行了比较。
In this paper, a convergent scheme of partitioning the wind field into divergent and rotational components over a limited domain is introduced and compared with some methods in current use.
一个好的和实用的差分格式应满足相容性、稳定性和收敛性。
A good and feasible difference scheme should satisfy the consistency, stability and convergence.
对该算法进行了推导,并给出了参数误差的差分方程,在理论上证明了算法的收敛性。
The algorithm is deduced and difference equation of parameter error given by which the algorithm convergence is proved theoretically.
鉴于BP网络存在着学习过程收敛速度慢、网络容错能力差的缺点,本文提出一种自构神经网络算法。
Self-configuring neural network is used because BP neural network has both slow convergence in learning course and poor fault tolerant ability.
该方法用数字差分滤波对采样信号进行处理,提高了信号中的定时分量,从而加快了算法的收敛速度。
In this algorithm, the digital differential filtering (DDF) is used to process the sampled signals, the timing component of signal is improved, and therefore the convergence speed is quickened.
通过理论分析,证明了闭环系统是半全局一致终结有界的,跟踪误差收敛到一个小的残差集内。
By theoretical(analysis, ) the closed-loop control system is proven to be semiglobally uniformly ultimately bounded, (with) tracking error converging to a residual set.
为提高计算效率,采用了多重网格法、当地时间步长和残差光顺等加速收敛措施。
Residual smoothing, local time stepping and a new multi-grid scheme were employed to improve calculative efficiency and accelerate the convergence.
还进一步讨论了B值鞅差序列随机加权和的收敛性。
Moreover, we discuss the convergence for randomly weighted sums of B-valued martingale difference series.
在全局对齐中,则是使用基于光差的全局优化方法在子像素级进行精确对齐,由于初值准确,因此全局算法能迅速收敛。
In the process of global alignment, because the initial value generated by the local alignment is close to the optimum one, the iterated algorithm could converge quickly.
采用了体积守恒形式的连续性方程,以防止由于气液两相密度差大而引起的数值求解中的不收敛现象。
The continuum equation with conserved volume type was used to avoid divergence because of the big difference in density between gas and liquid.
其次,针对解决此问题,使用传统分解协调法(DCM)的协调参数收敛性差的问题,应用遗传算法(GA)设计了协调参数。
Secondly, poor convergence is solved by utilizing traditional decomposition and coordination method(DCM), the coordination values are designed by genetic algorithm(GA).
给出了同阶差分格式线性组合的性质,从而可设计出更多收敛的多参数曲线细分格式。
Furthermore, the property of the linear combination of the difference schemes of the same order is analyzed, which can help to devise more convergent curve subdivision schemes with several parameters.
主要构造了求解一维隐式差分方程的四点嵌套迭代并行算法,并证明了它的收敛性。
The focal point is to construct a new embedding overlapped iterative algorithm for solving one-dimensional implicit difference equations.
传统的强化学习模型在整个学习过程中使用恒定学习速率,导致在未知环境下收敛速度慢且适应性差。
The learning process use the constant learning rate in the traditional reinforce learning model, because of that robot learn in a low convergence speed and with the poor adaptation.
为了求解非线性差分格式,本文建立一种加速单调迭代算法,并给出精确的收敛率估计。
To solve the nonlinear finite difference scheme, an accelerated monotone iterative method is presented, and the explicit estimate for the rate of convergence is given.
本文通过计算实例讨论了时域有限差分法的收敛性问题。
This paper discusses the convergence problem of the finite-difference time-domain (fd-td) method through some computing examples.
随机延迟微分方程数值方法中欧拉方法是唯一较为成熟、有效的方法,但欧拉方法的收敛性差,其收敛阶仅为二分之一。
Only the Euler method is popular and efficient among the numerical methods for the stochastic delay differential equations, but its order of convergence is only 1/2.
利用差商代替难以计算的精确导数,结合既约梯度法的思想建立新的算法;在目标函数一致凸的条件下证明了既约差商法的整体收敛性。
The difference coefficient was used to replace the exact derivative which is difficult to be computed, and a new algorithm was presented by using the idea of reduced gradient method.
本文用三维有限差分法,在计算电导率异常的电磁响应时作了一次网格收敛试验。
We present a mesh convergence test of the 3-d finite difference method for computing the electromagnetic response of a buried conductivity anomaly.
通过采用当地时间步和引入余差光滑处理法,进一步提高了格式的收敛性和稳定性。
The convergency and the stability of the scheme can be further improved by using local time stepping and residual smoothing.
通过采用当地时间步和引入余差光滑处理法,进一步提高了格式的收敛性和稳定性。
The convergency and the stability of the scheme can be further improved by using local time stepping and residual smoothing.
应用推荐