A Neural Network-Based Policy Iteration Algorithm with Global $$H^2$$ -Superlinear Convergence for Stochastic Games on Domains

In this work, we propose a class of numerical schemes for solving semilinear Hamilton–Jacobi–Bellman–Isaacs (HJBI) boundary value problems which arise naturally from exit time problems of diffusion processes with controlled drift. We exploit policy iteration to reduce the semilinear problem into a sequence of linear Dirichlet problems, which are subsequently approximated by a multilayer feedforward neural network ansatz. We establish that the numerical solutions converge globally in the \(H^2\)-norm and further demonstrate that this convergence is superlinear, by interpreting the algorithm as an inexact Newton iteration for the HJBI equation. Moreover, we construct the optimal feedback controls from the numerical value functions and deduce convergence. The numerical schemes and convergence results are then extended to oblique derivative boundary conditions. Numerical experiments on the stochastic Zermelo navigation problem are presented to illustrate the theoretical results and to demonstrate the effectiveness of the method.

     

一类非线性Schr(o)dinger方程的守恒差分法与Fourier谱方法

考察了一类带导数项的非线性Schrodinger方程的周期边值问题，提出了一种守恒的差分格式，在空间方向上采用Fourier谱方法，证明了格式的稳定性和收敛性．数值试验得到了与理论分析一致的结果．     

An optodynamic method for the real-time determination of the depth of a laser-drilled hole

An experimental method for determining the real-time depth of laser-drilled holes is presented. The proposed method involves detecting the laser-induced optoacoustic waves generated during the interaction of the laser beam with the material. Our optodynamic study involved measuring the propagation times of these waves as they traveled through the material and analyzing their temporal behavior during the drilling process. The experimental observations revealed an exponential relationship between the propagation time of the longitudinal stress wave and the number of consecutive laser pulses. Received: 25 October 2001 / Accepted: 27 October 2001 / Published online: 20 December 2001     

一类具有非局部反应扩散方程的奇摄动问题

本文讨论一类具非局部反应扩散方程的奇摄动初始边值问题，利用迭代法及微分不等式，研究了初始边值问题的存在唯一及其渐近性态。     

α－萘甲酰呋喃甲酰甲烷的合成及谱学性质的研究

本文用Ｃｌａｉｓｅｎ缩合反应合成了未见文献报道的一种新的β－二酮，α－萘甲酰呋喃甲酰甲烷（α－ＮＦＭ），考察了该化合物的红外光谱、质谱及核磁共振氢谱。确定其存在着互变异构。     

具有非局部边界条件的奇摄动反应扩散问题

本文研究了一类具有非局部边界条件的奇摄动反应扩散初始边值问题．在适当的条件下，利用比较定理讨论了问题解的渐近性态．     

基于反向传播神经网络的阻抗断层图像重建新方法

电阻抗成像是通过对物体表面电压、电流的测量来重建物体内部阻抗分布或变化图像的一种新颖计算机断层成像技术。阻抗断层图像重建是一种病态的、非线性的逆问题。提出了一种全新的阻抗断层图像重建方法，它利用反向传播神经网络来表征物体内部阻抗变化位置与物体表面电压变化大小的非线性映射关系，从而可以根据对物体表面测量电压的变化先准确定位阻抗变化区域，再用线性近似方法重建阻抗变化图像，这种方法不仅具有一定的抗噪能力，而且成像精度和空间分辨率都大大好于逆投影方法。     

A Strongly and Superlinearly Convergent SQP Algorithm for Optimization Problems with Linear Complementarity Constraints

This paper discusses a kind of optimization problem with linear complementarity constraints, and presents a sequential quadratic programming (SQP) algorithm for solving a stationary point of the problem. The algorithm is a modification of the SQP algorithm proposed by Fukushima et al. [Computational Optimization and Applications, 10 (1998), 5-34], and is based on a reformulation of complementarity condition as a system of linear equations. At each iteration, one quadratic programming and one system of equations needs to be solved, and a curve search is used to yield the step size. Under some appropriate assumptions, including the lower-level strict complementarity, but without the upper-level strict complementarity for the inequality constraints, the algorithm is proved to possess strong convergence and superlinear convergence. Some preliminary numerical results are reported.