A simultaneous perturbation stochastic approximation enhanced teaching-learning based optimization

As a novel population-based optimization algorithm, Teaching-Learning based Optimization (TLBO) possesses the advantages of single tuned parameter and global fast coarse search capability, at the same time it risks being trapped in local optima for function optimization due to premature convergence. In this paper, a TLBO-based memetic algorithm, namely TLBO-SPSA, is proposed, in which simultaneous perturbation stochastic approximation (SPSA) is incorporated into the canonical TLBO to enhance its local search capability and balance the global exploration and local exploitation as well. Numerical results on six well-known benchmark problems and comparisons with the state-of-the-art algorithms, e.g., conventional TLBO, I-TLBO, LGMS-FFO, IFFO, show that the proposed TLBO-SPSA outperforms the other algorithms in terms of the accuracy and convergence rate, especially the superior performance and robustness for solving higher dimensional problems.