Abstract
We introduce geomstats, a python package that performs computations onmanifolds such as hyperspheres, hyperbolic spaces, spaces of symmetric positivedefinite matrices and Lie groups of transformations. We provide efficient andextensively unit-tested implementations of these manifolds, together withuseful Riemannian metrics and associated Exponential and Logarithm maps. Thecorresponding geodesic distances provide a range of intuitive choices ofMachine Learning loss functions. We also give the corresponding Riemanniangradients. The operations implemented in geomstats are available with differentcomputing backends such as numpy, tensorflow and keras. We have enabled GPUimplementation and integrated geomstats manifold computations into keras deeplearning framework. This paper also presents a review of manifolds in machinelearning and an overview of the geomstats package with examples demonstratingits use for efficient and user-friendly Riemannian geometry.