Commit ab1cb25e authored by Johannes Pfeifer's avatar Johannes Pfeifer
Browse files

Filter out cases where invalid starting value leads to NaN or Inf in Jacobian.

Replaces cryptic message "Some element of Newton direction isn't finite. Jacobian maybe singular or there is a problem with initial values" with more expressive message.
parent 33053a73
...@@ -42,6 +42,24 @@ nn = size(x,1); ...@@ -42,6 +42,24 @@ nn = size(x,1);
% checking initial values % checking initial values
if jacobian_flag if jacobian_flag
[fvec,fjac] = feval(func,x,varargin{:}); [fvec,fjac] = feval(func,x,varargin{:});
if any(any(isinf(fjac) | isnan(fjac)))
[infrow,infcol]=find(isinf(fjac) | isnan(fjac));
M=evalin('base','M_'); %get variable names from workspace
fprintf('\nSTEADY: The Jacobian contains Inf or NaN. The problem arises from: \n\n')
for ii=1:length(infrow)
if infcol(ii)<=M.orig_endo_nbr
fprintf('STEADY: Derivative of Equation %d with respect to Variable %s (initial value of %s: %g) \n',infrow(ii),M.endo_names(infcol(ii),:),M.endo_names(infcol(ii),:),x(infcol(ii)))
else %auxiliary vars
fprintf('STEADY: Derivative of Equation %d with respect to Variable %s (initial value of %s: %g) \n',infrow(ii),M.endo_names(orig_var_index,:),M.endo_names(orig_var_index,:),x(infcol(ii)))
fprintf('\nSTEADY: The problem most often occurs, because a variable with\n')
fprintf('STEADY: exponent smaller than 1 has been initialized to 0. Taking the derivative\n')
fprintf('STEADY: and evaluating it at the steady state then results in a division by 0.\n')
error('An element of the Jacobian is not finite or NaN')
else else
fvec = feval(func,x,varargin{:}); fvec = feval(func,x,varargin{:});
fjac = zeros(nn,nn) ; fjac = zeros(nn,nn) ;
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment