## Copyright (C) 2008 Jaroslav Hajek, Marco Caliari
##
## This file is part of Octave.
##
## Octave is free software; you can redistribute it and/or modify it
## under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 3 of the License, or (at
## your option) any later version.
##
## Octave is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
## General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with Octave; see the file COPYING.  If not, see
## <http://www.gnu.org/licenses/>.

##-*- texinfo -*-
## @deftypefn {Function File} {} expm (@var{a})
## Return the exponential of a matrix, defined as the infinite Taylor
## series
## @iftex
## @tex
## $$
##  \exp (A) = I + A + {A^2 \over 2!} + {A^3 \over 3!} + \cdots
## $$
## @end tex
## @end iftex
## @ifinfo
## 
## @example
## expm(a) = I + a + a^2/2! + a^3/3! + ...
## @end example
## 
## @end ifinfo
## The Taylor series is @emph{not} the way to compute the matrix
## exponential; see Moler and Van Loan, @cite{Nineteen Dubious Ways to
## Compute the Exponential of a Matrix}, SIAM Review, 1978.  This routine
## uses Ward's diagonal
## @iftex
## @tex
## Pad\'e
## @end tex
## @end iftex
## @ifinfo
## Pade'
## @end ifinfo
## approximation method with three step preconditioning (SIAM Journal on
## Numerical Analysis, 1977).  Diagonal
## @iftex
## @tex
## Pad\'e
## @end tex
## @end iftex
## @ifinfo
## Pade'
## @end ifinfo
##  approximations are rational polynomials of matrices
## @iftex
## @tex
## $D_q(a)^{-1}N_q(a)$
## @end tex
## @end iftex
## @ifinfo
## 
## @example
##      -1
## D (a)   N (a)
## @end example
## 
## @end ifinfo
##  whose Taylor series matches the first
## @iftex
## @tex
## $2 q + 1 $
## @end tex
## @end iftex
## @ifinfo
## @code{2q+1}
## @end ifinfo
## terms of the Taylor series above; direct evaluation of the Taylor series
## (with the same preconditioning steps) may be desirable in lieu of the
## @iftex
## @tex
## Pad\'e
## @end tex
## @end iftex
## @ifinfo
## Pade'
## @end ifinfo
## approximation when
## @iftex
## @tex
## $D_q(a)$
## @end tex
## @end iftex
## @ifinfo
## @code{Dq(a)}
## @end ifinfo
## is ill-conditioned.
## @end deftypefn

function r = expm (a)

  if (! ismatrix (a) || ! issquare (a))
    error ("expm requires a square matrix")
  endif

  n = rows (a);
  # trace reduction
  a(a == -Inf) = -realmax;
  trshift = trace (a) / length (a);
  if (trshift > 0)
    a -= trshift*eye (n);
  endif
  # balancing
  [p, d, aa] = balance (a);
  # FIXME: can we both permute and scale at once? Or should we rather do this:
  # [p, xx, aa] = balance (a, "noscal");
  # [xx, d, aa] = balance (aa, "noperm");
  [f, e] = log2 (norm (aa, "inf"));
  s = max (0, e);
  s = min (s, 1023);
  aa *= 2^(-s);

  # Pade approximation for exp(A)
  c = [5.0000000000000000e-1,...
       1.1666666666666667e-1,...
       1.6666666666666667e-2,...
       1.6025641025641026e-3,...
       1.0683760683760684e-4,...
       4.8562548562548563e-6,...
       1.3875013875013875e-7,...
       1.9270852604185938e-9];

  a2 = aa^2;
  id = eye (n);
  x = (((c(8) * a2 + c(6) * id) * a2 + c(4) * id) * a2 + c(2) * id) * a2 + id;
  y = (((c(7) * a2 + c(5) * id) * a2 + c(3) * id) * a2 + c(1) * id) * aa;

  r = (x - y) \ (x + y);

  # Undo scaling by repeated squaring
  for k = 1:s 
    r ^= 2;
  endfor

  # inverse balancing
  ds = diag (s);
  r = ds * r / ds;
  r = r(p, p);
  # inverse trace reduction
  if (trshift >0)
    r *= exp (trshift);
  endif

endfunction