Artificial Neural Networks (ANN) are a powerful technique inspired by the human brain's nervous system and are widely employed as a crucial data processing tool today. The most important and difficult phase of an ANN is the training process, where the network's weights are optimized. As the number of connections in the neural network increases, so does the complexity of the weight optimization problem. Numerous algorithms and methods have been suggested over time to address this challenge. In recent years, one of the prominent techniques used for ANN training is meta-heuristic algorithms. This study evaluates the performance of several meta-heuristic algorithms for the solution of this problem. Specifically, six different algorithms, including the Grasshopper Optimization Algorithm, Artificial Hummingbird Algorithm, Arithmetic Optimization Algorithm, Crayfish Optimization Algorithm, Artificial Bee Colony and Tree-seed Algorithm were tested on 21 distinct datasets for ANN training. The performances of the algorithms were measured by using four popular metrics: precision, specificity, F1-score and sensitivity. The experimental findings reveal that the tested algorithms, particularly GOA, demonstrated high effectiveness in ANN training compared to the others. GOA produced the best results in 14 out of 21 datasets, achieving the top position in terms of average ranking success. These outcomes indicate that meta-heuristic algorithms offer a robust solution for handling the complex weight update processes in ANN.