Accelerating Vegetation Evolution with Gradient Descent and Martingale Strategies
Fei Peng, Rui Zhong, Chao Zhang, Jun Yu · 2024
In this paper, we propose two strategies to improve the vegetation evolution (VEGE) algorithm. The first strategy uses local gradient descent to accelerate the local search efficiency, while the second strategy employs the martingale method to enhance the optimization capability during the later stages of evolution. To validate the effectiveness of our proposals, we conducted independent experiments with 30 repetitions on the CEC 2017 test suite, comparing our proposal (VEGE + Strategy 1 + Strategy 2), VEGE + Strategy 1, VEGE + Strategy 2, and the original VEGE. The experiments were performed in both 10-dimensional and 30-dimensional cases, and the Holm multiple test was used to verify the significance of the differences. The experimental results’s convergence plots demonstrate that our proposal achieves faster convergence speeds and higher optimization accuracy.