The Gaussian kernel on the circle and spaces that admit isometric embeddings of the circle
This is an incremental result that clarifies limitations for kernel methods in machine learning on certain metric spaces.
The authors tackled the problem of whether the Gaussian kernel is positive definite on non-Euclidean spaces, specifically the circle, and found that it is not positive definite for any parameter, extending this result to spaces like spheres and Grassmannians.
On Euclidean spaces, the Gaussian kernel is one of the most widely used kernels in applications. It has also been used on non-Euclidean spaces, where it is known that there may be (and often are) scale parameters for which it is not positive definite. Hope remains that this kernel is positive definite for many choices of parameter. However, we show that the Gaussian kernel is not positive definite on the circle for any choice of parameter. This implies that on metric spaces in which the circle can be isometrically embedded, such as spheres, projective spaces and Grassmannians, the Gaussian kernel is not positive definite for any parameter.