at time t (even with the homotopic steps) then the size of the structural innovations is
reduced (at time t *only*). The advantage of this trick is that the seed of the random
number generator is not affected. The problem is that the structural innovations are now
heteroscedastic. Should add something to evaluate the importance of this heteroscedasticity problem.
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3346 ac1d8469-bf42-47a9-8791-bf33cf982152
* Fixed the evaluation of the distance to the deterministic steady
state in perfect_foresight_simulation.m (when options_.terminal_condition==2).
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3299 ac1d8469-bf42-47a9-8791-bf33cf982152
algorithm) with the solution of the linearized model.
* Bug fixes
* Cosmetic changes.
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3295 ac1d8469-bf42-47a9-8791-bf33cf982152
* Break Newton loop if err is NaN or if err is increasing (save useless iterations).
git-svn-id: https://www.dynare.org/svn/dynare/trunk@3152 ac1d8469-bf42-47a9-8791-bf33cf982152