Abstract:Symmetric Positive Definite (SPD) matrices have been widely used for data representation in many visual recognition tasks. The success mainly attributes to learning discriminative SPD matrices with encoding the Riemannian geometry of the underlying SPD manifold. In this paper, we propose a geometry-aware SPD similarity learning (SPDSL) framework to learn discriminative SPD features by directly pursuing manifold-manifold transformation matrix of column full-rank. Specifically, by exploiting the Riemannian geometry of the manifold of fixed-rank Positive Semidefinite (PSD) matrices, we present a new solution to reduce optimizing over the space of column full-rank transformation matrices to optimizing on the PSD manifold which has a well-established Riemannian structure. Under this solution, we exploit a new supervised SPD similarity learning technique to learn the transformation by regressing the similarities of selected SPD data pairs to their ground-truth similarities on the target SPD manifold. To optimize the proposed objective function, we further derive an algorithm on the PSD manifold. Evaluations on three visual classification tasks show the advantages of the proposed approach over the existing SPD-based discriminant learning methods.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve the problem of how to learn more discriminative SPD (symmetric positive definite) features while maintaining the Riemannian geometric structure of SPD matrices in order to improve performance in visual recognition tasks. Specifically, the author proposes a geometric - aware SPD similarity learning framework (SPDSL), directly pursuing a full - column - rank transformation matrix from the original SPD manifold to a more discriminative SPD manifold. #### Main challenges: 1. **Non - Euclidean data structure**: SPD matrices have a specific Riemannian manifold structure. Direct application of Euclidean geometry will lead to performance degradation and adverse effects, such as diffusion tensor dilation. 2. **Limitations of existing methods**: - **Flattening methods**: Mapping the SPD manifold to Euclidean space through tangent - space approximation will distort the original geometric structure. - **High - dimensional kernel methods**: Mapping the SPD manifold to a high - dimensional reproducing kernel Hilbert space (RKHS) will also introduce geometric distortion. - **Full - column - rank transformation matrix**: Directly learning a full - column - rank transformation matrix is difficult and the optimization process is complex. #### Solutions: 1. **Utilize the Riemannian geometry of the fixed - rank positive semi - definite (PSD) manifold**: A new method is proposed to transform the optimization problem from the space of the full - column - rank transformation matrix to the PSD manifold, thereby simplifying the optimization process and maintaining the geometric structure. 2. **Supervised SPD similarity learning**: Learn the transformation matrix by regressing the similarity of selected SPD data pairs to the true similarity on the target SPD manifold. 3. **Optimization algorithm**: Derive an optimization method based on the Riemannian conjugate gradient (RCG) algorithm to optimize the proposed objective function. #### Goals: - Propose a new geometric - aware SPD similarity learning framework (SPDSL) that can learn more discriminative SPD features while maintaining the Riemannian geometric structure. - Verify the effectiveness of the proposed method in three visual classification tasks (face recognition, material classification, and action recognition) and show its superiority over existing SPD - based discriminative learning methods. Through these improvements, the paper aims to overcome the limitations of existing methods and provide a more effective method for handling the application of SPD matrices in visual recognition tasks.

Geometry-aware Similarity Learning on SPD Manifolds for Visual Recognition

Learning Visually Aligned Semantic Graph for Cross-Modal Manifold Matching.

Learning Exemplar-Represented Manifolds in Latent Space for Classification.

SPD Manifold Deep Metric Learning for Image Set Classification

Dimensionality Reduction on SPD Manifolds: The Emergence of Geometry-Aware Methods

Learning a discriminative SPD manifold neural network for image set classification

Online Dictionary Learning On Symmetric Positive Definite Manifolds With Vision Applications

Deep Metric Learning on the SPD Manifold for Image Set Classification

Revisiting Metric Learning for SPD Matrix Based Visual Representation.

Riemannian Self-Attention Mechanism for SPD Networks

Locality preserving projection on SPD matrix Lie group: algorithm and analysis

Semantic And Neighborhood Preserving Dictionary Learning For Symmetric Positive-Definite Matrices

Deep Manifold Computing and Visualization Using Elastic Locally Isometric Smoothness

U-SPDNet: An SPD manifold learning-based neural network for visual classification

SymNet: A Simple Symmetric Positive Definite Manifold Deep Learning Method for Image Set Classification

Learning a Robust Representation via a Deep Network on Symmetric Positive Definite Manifolds

Riemannian Metric Learning for Symmetric Positive Definite Matrices

Manifold Kernel Sparse Representation of Symmetric Positive-Definite Matrices and Its Applications

2D-DLPP Algorithm Based on SPD Manifold Tangent Space.

Kernel Methods on the Riemannian Manifold of Symmetric Positive Definite Matrices

Riemannian Multinomial Logistics Regression for SPD Neural Networks