Distance functions and statistics. (Q1867203)

On a Riemannian manifold \((M, g)\) the class of \(R^\infty\) distance functions associated with the metric tensor \(g\) is considered. It is shown that the standard Riemannian distance function gives an upper bound for this class. Then, the paper studies the approximation of a sequence of points, \(p_1, \ldots, p_m\), on \(M\) by points \(\gamma_v(t_1), \ldots, \gamma_v(t_m)\) on the geodesic curve \(\gamma_v\) defined by \(v \in TM\). Assuming that \((M, g)\) is a complete smooth Riemannian surface with injectivity radius equal to infinity, it is shown that, for given \(p_i\) and \(t_i\), there exists a \(v\) which minimizes the least squares functional \(L= \sum_{i=1}^m(d(p_i, \gamma_v(t_i))^2\). Moreover, when the curvature and curvature change on \(M\) are small this minimum is unique.