TL;DR该论文提出了一种基于 Riemann 度量、Riemann 均值和 Riemann 优化的 SPD 流形自注意机制,用于改善所生成的深度结构表示的区分度,实验结果表明,该方法进一步减轻了信息退化问题并提高了准确性。
Abstract
Symmetric positive definite (SPD) matrix has been demonstrated to be an
effective feature descriptor in many scientific areas, as it can encode
spatiotemporal statistics of the data adequately on a curved Riemannian
manifold, i.e., spd manifold. Although there are many different ways t
使用正定对称 (SPD) 矩阵表示图像和视频,并考虑到所得空间的里曼尼几何,已被证明在许多识别任务中有益。本文引入了一种方法来构建一个更具判别力的低维 SPD 流形以处理高维 SPD 矩阵,并将学习表述为 Grassmann 流形上的优化问题。实验表明,与现有技术相比,我们的方法可使分类准确性显著提高。