A Stochastic Approximation Framework for a Class of Randomized Optimization Algorithms
A Stochastic Approximation Framework for a Class of Randomized Optimization Algorithms
复制标题
DOI：
10.1109/tac.2011.2158128
复制
发表时间：
2012
影响因子：
6.8
通讯作者：
Jiaqiao Hu;Ping Hu;H. Chang
中科院分区：
文献类型：
--
作者：
Jiaqiao Hu;Ping Hu;H. Chang
研究了一类基于随机抽样的求解一般不可微优化问题的算法。这些都是迭代方法，其基础是从可行解集合中采样并更新底层分布函数。特别是，我们提出了一个新的和系统的框架，调查这些算法的收敛性和渐近收敛速度，利用他们的连接到著名的随机逼近（SA）方法。这样的SA框架统一了我们对这些随机算法的理解，并为它们的设计和实现问题提供了新的见解。我们初步的数值实验表明，这些算法的新实现的基础上提出的框架可能会导致现有的程序，以提高性能。
We study a class of random sampling-based algorithms for solving general non-differentiable optimization problems. These are iterative approaches that are based on sampling from and updating an underlying distribution function over the set of feasible solutions. In particular, we propose a novel and systematic framework to investigate the convergence and asymptotic convergence rates of these algorithms by exploiting their connections to the well-known stochastic approximation (SA) method. Such an SA framework unifies our understanding of these randomized algorithms and provides new insight into their design and implementation issues. Our preliminary numerical experiments indicate that new implementations of these algorithms based on the proposed framework may lead to improved performance over existing procedures.