Lanczos' tau method for system of linear ODEs.

   a = tausys(x, y, ode_system, conditions, varargin)

   a = tausys(varargin) returns the coefficients, on basis P, of the 
   (n-1)th degree polinomial approximation yn = Pa, of the
   linear differential system dy/dx=F(t,y). Similar to
   tauode but for systems of odes.

Inputs (required)
   x              = independent tau variable (itau object).
   y              = dependent tau variable (dtau oject).
   ode_system     = system of odes (cell of char).
   conditions     = problem conditions (cell of char).

Inputs (optional)
   exact_solution = exact solution (cell of char).
   step           = step on the x vector to show the results.
   precond        = preconditioner ('no', 'ilu', 'diag').
                    for 'ndiad' define: 'numbd' (number od diagonals);
                    for 'ilu' define: 'milu', 'typeilu', 'droptol', 
                                      'thresh' and 'udiag'
   solver         = linear system solver.
   apsol          = boolean varargin to show graphically the solution.
   resid          = boolean varargin to show graphically the error.
   coeff          = boolean varargin to show the coefficients a.
   spy            = boolean varargin to spy the T matrix.
   saves          = name varargin to save the results at .mat.

   a              = approximate solution coefficients at basis P.

   [x, y] = tau('LegendreP', [-1 1], 10);
   a = tausys(x, y, {'diff(y2)+x^1*y1-y2 = 4*x^3-11*x^1+8*x^0'; ...
       'diff(y1)-y2 = 0'}, {'y1(-.5)=-2';'y2(-.5)=-4'});

See also
  tau, taupw, tausyspw and schursolver.