A trust region algorithm for parametric curve and surface fitting (Q1815878)

Let a family of curves or surfaces be given in parametric form \(x=f(s,\beta)\), where \(x\in\mathbb{R}^n\), \(\beta\in\mathbb{R}^m\), \(s\in S\subset\mathbb{R}^d\), \(d<n\). An algorithm is given to solve the problem: Find a shape vector \(\beta^*\) such that the manifold \(M^*=\{f(s,\beta^*)\}\) is a best fit to scattered data \(\{z_i\}^N_{i=1}\subset\mathbb{R}^n\). The set \(S\) may be all of \(\mathbb{R}^d\) or a closed, convex subset. An important feature is that the least distance part of the algorithm -- determination of the closest points for fixed \(\beta\) -- is uncoupled from the adjustment of the shape parameter \(\beta\). So the algorithm consists of two iterative processes. As an application a hyperboloid is fit to 63 data points given as triplets in \(\mathbb{R}^3\). So the shape vector has 9 components.