GRVFL-MV: Graph Random Vector Functional Link Based on Multi-View Learning

TL;DR

GRVFL-MV tackles RVFL's limited capacity to exploit multi-view data by fusing RVFL with MVL and GE. The model employs LFDA under GE with intrinsic and penalty graphs, and introduces a coupling term between two views to improve generalization, solved within a two-view optimization framework. Empirical results on UCI/KEEL, Corel5K, and AwA across high-dimensional, imbalanced, sparse, and various-dimensional datasets show statistically significant improvements over strong baselines, validated by Friedman and Nemenyi tests. The framework demonstrates robust, geometry-aware multi-view classification with practical impact for diverse data types and domains, while leaving room for extension to multiclass settings and additional views.

Abstract

The classification performance of the random vector functional link (RVFL), a randomized neural network, has been widely acknowledged. However, due to its shallow learning nature, RVFL often fails to consider all the relevant information available in a dataset. Additionally, it overlooks the geometrical properties of the dataset. To address these limitations, a novel graph random vector functional link based on multi-view learning (GRVFL-MV) model is proposed. The proposed model is trained on multiple views, incorporating the concept of multiview learning (MVL), and it also incorporates the geometrical properties of all the views using the graph embedding (GE) framework. The fusion of RVFL networks, MVL, and GE framework enables our proposed model to achieve the following: i) efficient learning: by leveraging the topology of RVFL, our proposed model can efficiently capture nonlinear relationships within the multi-view data, facilitating efficient and accurate predictions; ii) comprehensive representation: fusing information from diverse perspectives enhance the proposed model's ability to capture complex patterns and relationships within the data, thereby improving the model's overall generalization performance; and iii) structural awareness: by employing the GE framework, our proposed model leverages the original data distribution of the dataset by naturally exploiting both intrinsic and penalty subspace learning criteria. The evaluation of the proposed GRVFL-MV model on various datasets, including 27 UCI and KEEL datasets, 50 datasets from Corel5k, and 45 datasets from AwA, demonstrates its superior performance compared to baseline models. These results highlight the enhanced generalization capabilities of the proposed GRVFL-MV model across a diverse range of datasets.

Figures (6)

• Figure 1: The architecture of RVFL model.
• Figure 2: The effect of hyperparameter $(c_1, c_2)$ tuning on the accuracy (ACC) of some UCI and KEEL datasets on the performance of proposed GRVFL-MV model.
• Figure 3: The effect of coupling parameter $\rho$ tuning on the accuracy (ACC) of UCI and KEEl, AwA, and Corel5k datasets on the performance of proposed GRVFL-MV model.
• Figure 4: The effect of graph embedding parameter $\theta$ tuning on the accuracy (ACC) of Corel5k and AwA datasets on the performance of proposed GRVFL-MV model.
• Figure 5: Accuracy of Different Models on Highly Imbalanced Datasets.