It is considered t the Takagi-Sugeno-Kang fuzzy neural network and its modern variations. The use of regularization, random exclusion of rules from the rule base allows solving the problem of excessive similarity of rules in the rule base. The use of batch normalization to increase the generalizing properties of the network allows to increase the accuracy of the model, while maintaining the possibility of interpreting the results, which is characteristic of fuzzy neural networks. It is proposed to use an ensemble of fuzzy neural networks to increase the generalizing capabilities of the network. Studies of the Takagi-Sugeno-Kang fuzzy neural network for the task of diagnosing the coronavirus disease show that the proposed model works well and allows to improve the result.

This paper investigates the use of vision transformers (ViT) for skin cancer classification tasks, compared to convolutional models. We propose a novel ViT architecture that effectively classifies skin cancer images. Our findings suggest that ViT models have the potential to outperform convolutional models, especially with larger datasets.