Machine Learning-Assisted Approach for Optimizing Step Size of Hill Climbing Algorithm

Research on machine learning-assisted metaheuristics is a new and promising area for developing better optimization methods. This paper aims to improve the hill climbing algorithm by estimating the optimal step distance in each iteration using a neural network. This requires building a machine learning model, generating training data, and training the model. This model was then used in an alternating way with the traditional method. The climbing algorithm operates by searching for the best step size value in a few iterations and then using the step size suggested by the neural network in a number of subsequent iterations. The results show that a similar fitness/cost value can be obtained in this way as the most resource-intensive strategy, while the number of fitness calculations is halved.