Meta-heuristic optimization algorithms need a delicate balance between exploration and exploitation to search for global optima without premature convergence effectively. Parallel Sub-Class Modified Teac hing-learning-based optimization (PSC-MTLBO) is an improved version of TLBO proposed in this study to enhance search efficiency and solution accuracy. The proposed approach integrates three existing modifications-adaptive teaching factors, tutorial-based learning, and self-motivated learning-while introducing two novel enhancements: a sub-class division strategy and a challenger learners' model to enhance diversity and convergence speed. The proposed method was evaluated using three benchmark function sets (23 classical functions, 25 CEC2005 functions, and 30 CEC2014 functions) and two real-world truss topology optimization problems. Experimental results confirm that PSC-MTLBO performs better than normal TLBO, MTLBO, and other meta-heuristics such as PSO, DE, and GWO. For instance, PSC-MTLBO obtained the maximum overall rank in 80% of the test functions with the minimization of function errors by as much as 95% over traditional TLBO. In truss topology optimization, PSC-MTLBO designed lighter and more cost-effective structures with a weight reduction of 7.2% over the best solutions previously obtained. The challenger learners' model enhanced the adaptability, whereas the sub-class strategy increased the convergence and stability of results. In conclusion, PSC-MTLBO offers a remarkably efficient and scalable optimization framework and exhibits notable advances over current algorithms, with its suitability in solving complex optimization problems.
Tejani et al. (Fri,) studied this question.