Performance Analysis of Optimizer for Hyperparameter Tunning in Image Captioning

Hyper-parameter tuning is a critical component for optimizing the performance of deep learning models, including those used for image captioning. Hyper-parameters have a key impact in how accurate or coherent the produced captions are. This study explores the tuning performance of SSA, PSO, and GWO algorithms in the field of image captioning hyper-parameter optimization. This comparison is concerned with the convergence rates and the efficiency of tuning. The results show how well each optimizer is able to converge on the optimal set of hyper-parameter values, which in turn improves the quality of the captions generated. Such detailed analysis also brings forth the merits and demerits of each algorithm enabling the practitioners to choose the appropriate optimization algorithm for their machine learning application.