项目名称: 基于神经动态优化的一类伪凸优化问题研究
项目编号: No.61203299
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 自动化学科
项目作者: 彭勇刚
作者单位: 浙江大学
项目金额: 24万元
中文摘要: 优化是科学和工程中的常见问题,伪凸优化问题是一类重要的非凸优化问题,在科学研究和工程应用中具有重要应用。本课题研究基于神经动态优化方法的带约束伪凸优化问题,为伪凸优化问题的在线快速优化提供理论基础及方法。神经动态优化方法具有分布式、并行优化且可硬件实现的优点,课题采用对偶原理及优化问题KKT优化条件分析方法设计优化神经网络,并通过Lyapunov稳定性分析保证优化神经网络的稳定性,并快速收敛到优化问题的解,最后利用所提出的神经动态优化方法解决典型的实际工程伪凸优化问题- - 分数规划及伪凸二次规划问题。通过本课题的研究,将神经动态优化方法推广到非凸优化研究领域,为伪凸优化问题的神经动态优化提供理论基础和求解方法,并为解决其他在线非凸优化问题提供思路。
中文关键词: 神经动态优化;模型预测控制;神经网络;凸优化;
英文摘要: Optimization is an common problem in scientific and engineering fields.Pseudoconvex optimization is one of important nonconvex optimization problems and it has important applications in scientific research,engineering applications. Pseudoconvex optimization with constraints will be researched based on neurodynamical optimization method and it will provide theory fundation and method for online Pseudoconvex optimization problem. Neurodynamical optimization method has the advantage of inherent nature of parallel and distributed computing and it can be implemented in hardware.Dual principle and KKT condition analysis will be used to design optimization neural network. Then Lyapunov stability analysis method will be used to guarantee that optimization neural network is steady and it will quickly converge to solution of optimization probelm. Finally the neurodynamical optimization method proposed in this project will be used to solve typical practical engineering pseudoconvex optimization problems- - fractional programmig and pseudoconvex quadratic programming. Through the study in this project neurodynamical optimization method will be expanded to nonconvexity field and the results will provide theoretical foundation and solution methods for online nonconvexity optimization problems.
英文关键词: neurodynamical optimization;model predictive control;neural network;convex optimization;