基于元胞自动机的教与学优化算法

Cellular automaton-based teaching-learning optimization algorithm

为解决教与学优化(TLBO)算法易陷入局部最优的问题,提出了一种基于元胞自动机的教与学优化算法(CATLBO)。算法建立了四边形网状元胞自动机模型并指定其邻域结构和规则。为保持种群多样性,在教学阶段提出以一定的概率接收退步个体的策略;为加快收敛并保证解的精度,在学习阶段制定不同学习规则,劣势个体向优势个体学习,优势个体执行混沌扰动进行自我学习。使用多个Benchmark测试函数和经典TSP问题对算法进行了仿真。结果表明:CATLBO算法全局搜索能力强,与基本TLBO等算法相比,在处理高维多峰问题上更具优势。

To overcome the problem that teaching-learning-based optimization( TLBO) algorithm is prone to local optimum,a cellular automaton-based TLBO( CATLBO) algorithm is proposed. The algorithm establishes a quadrilateral mesh model for CA with specified neighborhood structure and rules. In order to maintain diversity of the population,a strategy is proposed to receive individuals who setback in teaching phase within a certain probability. To speed up the convergence and ensure precision of the solution,different learning rules are customized during the learning phase. Outstanding individuals perform chaotic disturbances for self-learning,and inferiority individuals learn from outstanding ones. Use multiple Benchmarks test functions and classical TSP problems to simulate on algorithm. The result show that the CATLBO algorithm has strong global searching ability and is superior to algorithms such as TLBO in dealing with high dimensional multi-peak problems.