1: \begin{abstract}

2:     We study first-order optimization algorithms for computing the barycenter of Gaussian distributions with respect to the optimal transport metric.

3:     Although the objective is geodesically non-convex,

4:     Riemannian GD empirically converges rapidly, in fact faster than off-the-shelf methods such as Euclidean GD and SDP solvers.

5:     This stands in stark contrast to the best-known theoretical results for Riemannian GD, which depend

6:     exponentially on the dimension. 

7:     In this work, we prove new geodesic convexity results which provide stronger control of the iterates, yielding

8:     a dimension-free convergence rate.

9:     Our techniques also enable the analysis of two related notions of averaging, the entropically-regularized barycenter and the geometric median, providing the first convergence guarantees for Riemannian GD for these problems.

10: \end{abstract}

11: