一种改进的GA+BP模糊逻辑系统混合学习算法

分析了遗传算法(GeneticAlgorithm)和BP算法在模糊逻辑系统参数寻优问题上的优缺点,提出一种基于改进的GA+BP模糊逻辑系统混合学习算法。该算法克服了由于学习率选取不当对整个遗传进化过程造成的不利影响,改进了遗传算子具体操作步骤,并对当前最优个体采用最优保留策略。充分利用了杂交、变异选择算子在全变量空间以较大概率搜索全局解的特点,以及在解点附近BP算子快速、精确地收敛的特点。仿真实验表明,改进算法与原算法相比,在满足同样精度的条件下,具有较快的收敛速度。

An Improved Hybrid-Learning Algorithm Based on GA+BP in Fuzzy Logical Systems

The advantages and disadvantages of genetic and BP network algorithm on parameter optimization of fuzzy logical systems were analyzed and an improved hybrid learning algorithm based on GA+BP in fuzzy logical systems was proposed. The new algorithm can overcome the adverse effects due to the unsuitable learning rate toward the whole genetic evolutionary process, improve specific operation steps of genetic operators, and adopt elitist strategy toward current optimization individual. The peculiarities which crossover, mutation and selection operators can search the whole variable space in high probability and BP operators can converge accurately and fast near the solution have been adopted sufficiently here. Simulation results demonstrate that the improved algorithm is more efficient to improve convergence speed than the standard algorithm in the condition of satisfying the same accuracy.