 This research compared two types of machine learning algorithms, convolutional neural network, CNN, and multi-layer perceptron mixer, MLP mixer. It found that the MLP mixer model had better performance than the CNN model on smaller datasets, but was less accurate on larger ones. To address this issue, the researchers used Bayesian deep learning, BDL, which uses variational inference and Monte Carlo dropout to reduce uncertainty in the model. They found that BDL improved the performance of the MLP mixer model on all datasets, improving it by up to 17%. Additionally, they found that the CNN model did not benefit from BDL, suggesting that the MLP mixer model may be more suitable for smaller datasets. This article was authored by Abdullah A. Abdullah, Masud M. Hassan, and Yasin T. Mustafa.