S2FunMLS.S2FunMLS edit page

A class representing a function on the 2-sphere S^2.

Syntax

S2F = S2FunMLS(nodes, values);
S2F = S2FunMLS(nodes, values, 'degree', 3, 'oF', 4);
S2F = S2FunMLS(nodes, values, 'delta', 5*degree, 'weight', @(t)(...));
S2F = S2FunMLS(nodes, values, 'centered', true, 'monomials', true, ...
  'subsample', 'tangent', true);
S2F = S2FunMLS(nodes, values, 'detectOutliers', ...
  'use_vor_weights', 'use_smooth_delta');

Input

nodes vector3d (data points)
values array of function values assigned to the nodes

Output

S2F S2FunMLS

Options

degree polynomial degree used for approximation
oF oversampling factor; nn is S2F.dim times this factor
oF_max maximal oversampling factor for range search
delta support radius of the weight function; delta = 0 uses KNN
monomials use a monomial basis, otherwise spherical harmonics
centered evaluate the basis in local coordinates around the north pole
tangent use monomials on the tangent plane (requires centered = true) (NOTE: centered and tangent automatically enable the monomial basis)
w @function_handle (weight function)
predefined choices include 'C1hat' (default), 'const', 'cos', 'hat', 'indicator', 'squared hat', 'wendland', and 'wendlandC6'
use_smooth_delta use a smooth local support radius with about S2F.nn neighbors at each center
use_vor_weights multiply the local weights by Voronoi areas
distance metric for neighbor search (default: 'euclidean')
s symmetry of the nodes
regularize use goal-oriented regularization of the local systems
mincond center-amplification threshold where regularization starts
maxcond center-amplification threshold where full correction is used
targetcond center amplification approached at full correction (The property names are retained for compatibility. They no longer refer to the ordinary condition number of the Gram matrix.)
outlierDetectionRange number of neighbors used for outlier detection

Flags

detectOutliers detect local outliers and reduce their MLS weights
subsample select a subset that minimizes the Lebesgue constant