GB-RVFL: Fusion of randomized neural network and granular ball computing

Abstract

The random vector functional link (RVFL) network is a prominent classification model with strong generalization ability. However, RVFL treats all samples uniformly, ignoring whether they are pure or noisy, and its scalability is limited due to the need for inverting the entire training matrix. To address these issues, we propose granular ball RVFL (GB-RVFL) model, which uses granular balls (GBs) as inputs instead of training samples. This approach enhances scalability by requiring only the inverse of the matrix of GBs’ centers and improves robustness against noise and outliers through the coarse granularity of GBs. Furthermore, RVFL overlooks the dataset's geometric structure. To address this, we propose graph embedding GB-RVFL (GE-GB-RVFL) model, which fuses granular computing and graph embedding (GE) to preserve the topological structure of GBs and enhances the interpretability of the proposed model. The proposed GB-RVFL and GE-GB-RVFL models are evaluated on KEEL, UCI, NDC and biomedical datasets such as Alzheimer's disease diagnosis and breast cancer prediction. The experimental evaluation demonstrates that the proposed models outperform baseline models in efficacy, robustness, scalability, and interpretability. The source codes of the proposed GB-RVFL and GE-GB-RVFL models are available at https://github.com/mtanveer1/GB-RVFL. © 2024 Elsevier Ltd