Home > manopt > manifolds > sphere > spheresymmetricfactory.m

spheresymmetricfactory

PURPOSE ^

Returns a manifold struct to optimize over unit-norm symmetric matrices.

SYNOPSIS ^

function M = spheresymmetricfactory(n)

DESCRIPTION ^

 Returns a manifold struct to optimize over unit-norm symmetric matrices.

 function M = spheresymmetricfactory(n)

 Manifold of n-by-n real symmetric matrices of unit Frobenius norm.
 The metric is such that the sphere is a Riemannian submanifold of the
 space of nxn symmetric matrices with the usual trace inner product, i.e.,
 the usual metric <A, B> = trace(A'*B).
 
 See also: spherefactory obliquefactory spherecomplexfactory

CROSS-REFERENCE INFORMATION ^

This function calls: This function is called by:

SUBFUNCTIONS ^

SOURCE CODE ^

0001 function M = spheresymmetricfactory(n)
0002 % Returns a manifold struct to optimize over unit-norm symmetric matrices.
0003 %
0004 % function M = spheresymmetricfactory(n)
0005 %
0006 % Manifold of n-by-n real symmetric matrices of unit Frobenius norm.
0007 % The metric is such that the sphere is a Riemannian submanifold of the
0008 % space of nxn symmetric matrices with the usual trace inner product, i.e.,
0009 % the usual metric <A, B> = trace(A'*B).
0010 %
0011 % See also: spherefactory obliquefactory spherecomplexfactory
0012 
0013 % This file is part of Manopt: www.manopt.org.
0014 % Original author: Nicolas Boumal, April 17, 2015.
0015 % Contributors:
0016 % Change log:
0017 %
0018 %   Oct. 8, 2016 (NB)
0019 %       Code for exponential was simplified to only treat the zero vector
0020 %       as a particular case.
0021 %
0022 %   Oct. 22, 2016 (NB)
0023 %       Distance function dist now significantly more accurate for points
0024 %       within 1e-7 and less from each other.
0025 %
0026 %   July 20, 2017 (NB)
0027 %       The distance function is now even more accurate.
0028 
0029 
0030     M.name = @() sprintf('Sphere of symmetric matrices of size %d', n);
0031     
0032     M.dim = @() n*(n+1)/2 - 1;
0033     
0034     M.inner = @(x, d1, d2) d1(:).'*d2(:);
0035     
0036     M.norm = @(x, d) norm(d, 'fro');
0037     
0038     M.dist = @(x, y) real(2*asin(.5*norm(x - y, 'fro')));
0039     
0040     M.typicaldist = @() pi;
0041     
0042     M.proj = @proj;
0043     function xdot = proj(x, d)
0044         d = (d+d.')/2;
0045         xdot = d - x*(x(:).'*d(:));
0046     end
0047     
0048     M.tangent = @proj;
0049     
0050     % For Riemannian submanifolds, converting a Euclidean gradient into a
0051     % Riemannian gradient amounts to an orthogonal projection.
0052     M.egrad2rgrad = @proj;
0053     
0054     M.ehess2rhess = @ehess2rhess;
0055     function rhess = ehess2rhess(x, egrad, ehess, u)
0056         % these are not explicitly required, given the use.
0057         % egrad = (egrad + egrad.')/2;
0058         % ehess = (ehess + ehess.')/2;
0059         rhess = proj(x, ehess) - (x(:)'*egrad(:))*u;
0060     end
0061     
0062     M.exp = @exponential;
0063     
0064     M.retr = @retraction;
0065 
0066     M.log = @logarithm;
0067     function v = logarithm(x1, x2)
0068         v = proj(x1, x2 - x1);
0069         di = M.dist(x1, x2);
0070         % If the two points are "far apart", correct the norm.
0071         if di > 1e-6
0072             nv = norm(v, 'fro');
0073             v = v * (di / nv);
0074         end
0075     end
0076     
0077     M.hash = @(x) ['z' hashmd5(x(:))];
0078     
0079     M.rand = @() random(n);
0080     
0081     M.randvec = @(x) randomvec(n, x);
0082     
0083     M.lincomb = @matrixlincomb;
0084     
0085     M.zerovec = @(x) zeros(n);
0086     
0087     M.transp = @(x1, x2, d) proj(x2, d);
0088     
0089     M.pairmean = @pairmean;
0090     function y = pairmean(x1, x2)
0091         y = x1+x2;
0092         y = y / norm(y, 'fro');
0093     end
0094 
0095     % TODO : check isometry and fix.
0096     M.vec = @(x, u_mat) u_mat(:);
0097     M.mat = @(x, u_vec) reshape(u_vec, [n, m]);
0098     M.vecmatareisometries = @() false;
0099 
0100 end
0101 
0102 % Exponential on the sphere
0103 function y = exponential(x, d, t)
0104 
0105     if nargin == 2
0106         % t = 1;
0107         td = d;
0108     else
0109         td = t*d;
0110     end
0111     
0112     nrm_td = norm(td, 'fro');
0113     
0114     if nrm_td > 0
0115         y = x*cos(nrm_td) + td*(sin(nrm_td)/nrm_td);
0116     else
0117         y = x;
0118     end
0119 
0120 end
0121 
0122 % Retraction on the sphere
0123 function y = retraction(x, d, t)
0124 
0125     if nargin == 2
0126         t = 1;
0127     end
0128     
0129     y = x + t*d;
0130     y = y / norm(y, 'fro');
0131 
0132 end
0133 
0134 % Uniform random sampling on the sphere.
0135 function x = random(n)
0136 
0137     x = randn(n);
0138     x = (x + x.')/2;
0139     x = x/norm(x, 'fro');
0140 
0141 end
0142 
0143 % Random normalized tangent vector at x.
0144 function d = randomvec(n, x)
0145 
0146     d = randn(n);
0147     d = (d + d.')/2;
0148     d = d - x*(x(:).'*d(:));
0149     d = d / norm(d, 'fro');
0150 
0151 end

Generated on Fri 08-Sep-2017 12:43:19 by m2html © 2005