Scalable Semi-Supervised Learning through Combined Anchor-based Graph and Flexible Manifold Embedding

Abstract

This paper focuses on graph-based semi-supervised learning, specifically for large-scale graphs used in inductive multi-class classification. The proposed method aims to overcome limitations in current scalable graph-based semi-supervised learning techniques. The key innovation is integrating the anchor graph calculation into the learning model, rather than treating it as a separate, offline step. This approach involves several essential tasks, including simultaneously estimating unlabeled samples, mapping the feature space to the label space, creating an affinity matrix for the anchor graph, and using labels and features associated with anchor points to construct the graph. The experimental results, conducted with large datasets, demonstrate a positive trend, showing higher accuracy and greater stability compared to existing scalable graph-based semi-supervised learning methods.

Fadi Dornaika

Ikerbasque Research Professor

Ikerbasque Research Professor with expertise in computer vision, machine learning, and pattern recognition.

Zoulfikar Ibrahim

Professor and Software Engineer

I develop scalable graph-based machine learning methods and teach cutting-edge technologies in web and software development. My focus spans semi-supervised learning, data analysis, and full-stack engineering.

Nagore Barrena

Assistant Professor

My research focuses on computer vision and deep learning, with applications to biomedical and natural images.