Intuitionistic Fuzzy Deep Random Vector Functional Link Networks for Robust Classification

There’s a new preprint on arXiv with the identifier 2608.10007 that introduces some exciting changes to deep Random Vector Functional Link networks. It presents two new models called IF-dRVFL and IF-edRVFL, which use intuitionistic fuzzy theory to improve how these networks handle noise and outliers during classification. By factoring in both membership and non-membership degrees, these models tap into neighborhood information in the kernel space. They determine membership based on how close samples are to their class centroids and use non-membership to show uncertainty. This approach addresses weaknesses in current deep randomized neural networks that struggle with noisy real-world data. It’s a noteworthy contribution to deep learning and fuzzy systems!