动机背景
针对单目标多峰优化问题,传统的优化算法面临着局部最优解陷阱的挑战,无法充分挖掘解空间中的潜在全局最优解。
为了克服这一问题,我们提出了一种基于黎曼流形的优化方法。
目的
该方法能够在高维、复杂的解空间中灵活地搜索,并通过黎曼度量保持了解空间的几何结构。这种方法不仅具有较强的全局搜索能力,还能够快速、精确地收敛到最优解。
方法
例子1:
Initial population(初始化种群)
Neighbourhood(找到每个点的邻域点)
Compare fitness(计算适应度值)(Delete the value with small fitness in the neighborhood.)
Category coloring to classification(得到分类结果)
First class: blue [1, 2, 4]
Second category: red [0, 3]
流程图
示例图
例子2
2D Gragh Distribution of Children Points when Generation is 5:
3D Gragh Distribution of Children Points when Generation is 5:
2D Gragh Distribution of Children Points when Generation is 10
3D Gragh Distribution of Children Points when Generation is 10:
结论
- Persistence-Based Clustering was proposed to enhance evolutionary algorithms.
- Developing the Chaotic Evolutionary Clustering Algorithm (CECA).
- Quickly and accurately converge to the optimal solution.