A Geometry Preserving Kernel over Riemannian Manifolds
نویسندگان
چکیده مقاله:
Abstract- Kernel trick and projection to tangent spaces are two choices for linearizing the data points lying on Riemannian manifolds. These approaches are used to provide the prerequisites for applying standard machine learning methods on Riemannian manifolds. Classical kernels implicitly project data to high dimensional feature space without considering the intrinsic geometry of data points. Projection to tangent spaces truly preserves topology along radial geodesics. In this paper, we propose a method for extrinsic inference on Riemannian manifold using kernel approach while topology of the entire dataset is preserved. We show that computing the Gramian matrix using geodesic distances, on a complete Riemannian manifold with unique minimizing geodesic between each pair of points, provides a feature mapping which preserves the topology of data points in the feature space. The proposed approach is evaluated on real datasets composed of EEG signals of patients with two different mental disorders, texture, visual object classes, and tracking datasets. To assess the effectiveness of our scheme, the extracted features are examined by other state-of-the-art techniques for extrinsic inference over symmetric positive definite (SPD) Riemannian manifold. Experimental results show the superior accuracy of the proposed approach over approaches which use kernel trick to compute similarity on SPD manifolds without considering the topology of dataset or partially preserving topology.
منابع مشابه
Riemannian manifolds in noncommutative geometry
We present a definition of Riemannian manifold in noncommutative geometry. Using products of unbounded Kasparov modules, we show one can obtain such Riemannian manifolds from noncommutative spinc manifolds; and conversely, in the presence of a spinc structure. We also show how to obtain an analogue of Kasparov's fundamental class for a Riemannian manifold, and the associated notion of Poincaré ...
متن کاملKernel density estimation on Riemannian manifolds
The estimation of the underlying probability density of n i.i.d. random objects on a compact Riemannian manifold without boundary is considered. The proposed methodology adapts the technique of kernel density estimation on Euclidean sample spaces to this non-Euclidean setting. Under sufficient regularity assumptions on the underlying density, L 2 convergence rates are obtained. Index Terms — No...
متن کاملThe Geometry of Sub-riemannian Three-manifolds
The local equivalence problem for sub-Riemannian structures on threemanifolds is solved. In the course of the solution, it is shown how to attach a canonical Riemannian metric and connection to the given sub-Riemannian metric and it is shown how all of the differential invariants of the sub-Riemannian structure can be calculated. The relation between the completeness of the sub-Riemannian metri...
متن کاملRiemannian Geometry over different Normed Division Algebras
We develop a unifed theory to study geometry of manifolds with different holonomy groups. They are classified by (1) real, complex, quaternion or octonion number they are defined over and (2) being special or not. Specialty is an orientation with respect to the corresponding normed algebra A. For example, special Riemannian A-manifolds are oriented Riemannian, Calabi-Yau, Hyperkähler and G2-man...
متن کاملSeparability preserving Dirac reductions of Poisson pencils on Riemannian manifolds
Dirac deformation of Poisson operators of arbitrary rank is considered. The question when Dirac reduction does not destroy linear Poisson pencils is studied. A class of separability preserving Dirac reductions in the corresponding quasi-bi-Hamiltonian systems of Benenti type is discussed. Two examples of such reductions are given. This paper will appear in J. Phys. A: Math. Gen. AMS 2000 Subjec...
متن کاملOptimized Kernel-based Projection Space of Riemannian Manifolds
Recent advances in computer vision suggest that encoding images through Symmetric Positive Definite (SPD) matrices can lead to increased classification performance. Taking into account manifold geometry is typically done via embedding the manifolds in tangent spaces, or Reproducing Kernel Hilbert Spaces (RKHS). Recently it was shown that projecting such manifolds into a kernel-based random proj...
متن کاملمنابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ذخیره در منابع من قبلا به منابع من ذحیره شده{@ msg_add @}
عنوان ژورنال
دوره 6 شماره 2
صفحات 321- 334
تاریخ انتشار 2018-07-01
با دنبال کردن یک ژورنال هنگامی که شماره جدید این ژورنال منتشر می شود به شما از طریق ایمیل اطلاع داده می شود.
میزبانی شده توسط پلتفرم ابری doprax.com
copyright © 2015-2023