Abstract
This chapter presents an innovative semi-supervised model that aims to improve data analysis by addressing the limitations of relying on single features. Instead, it advocates the simultaneous use of multiple features to improve performance and avoid misleading conclusions. This algorithm focuses on the joint construction of a consistent graph and label estimation. Key highlights of the proposed method include (1) multi-view data representation: the algorithm processes multi-view data, ensuring a richer and more comprehensive representation; (2) semi-supervised classification: by using both labeled and unlabeled data, the algorithm improves classification accuracy and generalization; and (3) improved visualization: in addition to classification, the method also focuses on improving data visualization to provide better insight and understanding. Experimental results on various image databases show the superiority of the fusion approach over single-feature methods or alternative fusion algorithms. These results underline the importance of leveraging multiple features while creating unified graphs to improve performance in semi-supervised data classification and visualization tasks.
Fadi Dornaika
Ikerbasque Research Professor
Ikerbasque Research Professor with expertise in computer vision, machine learning, and pattern recognition.
Nagore Barrena
Assistant Professor
My research focuses on computer vision and deep learning, with applications to biomedical and natural images.