基于变量转换的并行优化算法

Parallel Optimization Algorithm Based on Variable Transformation

针对大规模边界约束优化问题,现有并行变量转换(PVT)算法不适于直接求解。基于此,采用内点法和逐步下降的思想,提出一个并行求解边界约束最优化问题的可行算法。在下降方向满足梯度相关、步长满足Goldstein规则的条件下,证明该算法的收敛性。当约束失效时,该算法退化为求解无约束的PVT算法,从而成为原有算法向约束优化问题的一个推广。

The present Parallel Variable Transformation（PVT） algorithm is not suitable for large scale optimization problem with bounded constraints.This paper proposes a parallel feasible algorithm for bounded constrained optimization problem by the interior-point and the gradual reduction.Yet,the convergence of the algorithm is obtained under the conditions that the reduction direction is gradient dependent and the search step satisfies Goldstein criteria.When the constraints losing effectiveness,it deqenerates into the unconstrained PVT algorithm.It is an extension of unconstrained PVT algorithm.