Abstract:Graph-based solutions have achieved state-of-the-art performance on unsupervised multi-view feature selection. However, existing methods generally characterize the sample similarities first by constructing multiple fixed graphs with manually determined parameters, and then perform the feature selection on a composite one. They will suffer from two severe problems: (1) The fixed graphs may be unreliable as the raw multi-view features usually contain adverse noises and cannot accurately capture the intrinsic sample relations. (2) The graph construction and feature selection are separate and independent, the two-step learning may lead to sub-optimal performance. To tackle these problems, in this paper, we propose an effective unsupervised multi-view feature selection method, dubbed as Nonnegative Structured Graph Learning (NSGL). Specifically, we develop a unified learning framework, which directly learns the structured graph from the raw features by imposing a rank constraint, and simultaneously performs adaptive feature selection with exploiting the complementarity of multi-view features. Besides, we introduce the pseudo label learning to extract the discriminative semantic information in unsupervised scenarios and steer the graph learning process. The informative features are finally selected by forcing the feature selection matrix to be sparse in rows with sparse regression. To solve the challenging optimization problem, we first transform the formulated problem into an equivalent one that can be tackled more easily, and then develop an efficient alternate optimization algorithm guaranteed with convergence to calculate the solution iteratively. Extensive experiments on several widely tested benchmarks demonstrate the superiority of NSGL compared with several state-of-the-art approaches. (C) 2020 Elsevier B.V. All rights reserved.

Scalable graph based non-negative multi-view embedding for image ranking

Feature Extraction Via Multi-View Non-Negative Matrix Factorization with Local Graph Regularization.

Robust Multi-View Manifold Ranking For Image Retrieval

Sparse Multi-Graph Ranking Towards Social Image Retrieval

Multi-View Graph Embedding Learning for Image Co-Segmentation and Co-Localization

Multi-Manifold Ranking: Using Multiple Features for Better Image Retrieval.

ROMIR: Robust Multi-View Image Re-Ranking

Multi-view feature selection via Nonnegative Structured Graph Learning

Ranking Graph Embedding for Learning to Rerank

Visual Reranking through Weakly Supervised Multi-graph Learning

Three Tiers Neighborhood Graph and Multi-graph Fusion Ranking for Multi-feature Image Retrieval: A Manifold Aspect.

Multi-view embedding learning via robust joint nonnegative matrix factorization

Social Neighborhood Graph and Multigraph Fusion Ranking for Multifeature Image Retrieval

From Document to Image: Learning a Scalable Ranking Model for Content Based Image Retrieval

Multi-View Embedding Learning for Incompletely Labeled Data.

Multi-View 3d Object Retrieval with Deep Embedding Network

Flexible Multi-View Unsupervised Graph Embedding

Multi-Manifold Sparse Graph Embedding for Multi-Modal Image Classification

Unsupervised Multi-view Nonlinear Graph Embedding.

Unified Hypergraph For Image Ranking In A Multimodal Context

Efficient Multi-Ranking Based On View Selection For Content Based Image Retrieval