A Comparison of the Pattern Search Algorithm and the Modified PATH Algorithm for Optimizing an Individual Tree Model
A Comparison of the Pattern Search Algorithm and the Modified PATH Algorithm for Optimizing an Individual Tree Model
复制标题
模式搜索算法与改进的路径算法优化单树模型的比较
DOI：
10.1093/forestscience/36.2.394
复制
发表时间：
1990
期刊：
影响因子：
1.4
通讯作者：
J. Brodie
中科院分区：
文献类型：
--
作者：
Atsushi Yoshimoto;Robert G. Haight;J. Brodie
开发了一种新的机架级动态规划算法，用于确定多个机架入口的最佳剩余直径分布。该算法由PATH算法以及区域限制策略和迭代动态规划的概念组成。采用所提出的算法，利用林分预测模型构建了动态规划模型，林分预测模型是落基山脉北部森林类型的单树/距离无关的生长模型。通过与非线性规划算法（Hooke 和 Jeeves 方法）的比较，讨论了该算法的收敛性、准确性和效率。对于具有三个或更多细化的问题，所提出的算法比 Hooke 和 Jeeves 方法用更少的计算时间产生了更好的解决方案。对于一、二细化问题，Hooke 和 Jeeves 方法提供了更好的解决方案。所提出算法的优点源于动态规划方法能够避免在解决方案中包含多个部分局部最优，以及与凹优化的既定原理有更清晰的关系。为了。科学。 36（2）：394-412。
A new stand-level dynamic programming algorithm was developed for determining the optimal residual diameter distribution for multiple stand entries. The proposed algorithm consists of the PATH algorithm and the concept of region-limiting strategies and iterated dynamic programming. Employing the proposed algorithm, a dynamic programming model was constructed with the Stand Prognosis Model, a single-tree/distance-independent growth model for forest types in the northern Rocky Mountains. Convergence, accuracy, and efficiency of the algorithm are discussed in comparison with a nonlinear programming algorithm, the Hooke and Jeeves method. For problems with three or more thinnings, the proposed algorithm yielded superior solutions with less computation time than did the Hooke and Jeeves method. For one and two thinning problems, the Hooke and Jeeves method provided better solutions. The advantage of the proposed algorithm stems from the ability of dynamic programming approaches to avoid including multiple partial local optima in the solution, and a clearer relationship to established principles of concave optimization. For. Sci. 36(2):394-412.