Skin cancer is a significant public health concern, necessitating accurate and timely detection for effective treatment. Deep learning models have emerged as promising tools for skin cancer classification, demonstrating remarkable accuracy in various studies. However, the performance of deep learning models heavily relies on the choice of optimizer, which affects the training process and convergence speed. In this paper, we investigate the accuracy variance between different optimizers in deep learning models for skin cancer classification. Different deep learning architectures, such as convolutional neural networks (CNNs) are implemented on HAM 10000 (Human Against Machine) dataset to train the skin cancer classification models. The obtained results compare different optimizers such as Root Mean Square Propagation (RMSProp), Nadam, AdaDelta, Stochastic Gradient Descent (SGD), Adamax, Adagrad …