You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current algorithm of iDMRG followed by fDMRG can not easily converge for larger size system. This is because iDMRG is growing too fast, and the MPS generated by iDMRG on the fly is way deviated from the ground state, making the local optimization of fDMRG hard to converge. So we propose a new algorithm here. This is by first growing the system by 2-sites, using an iDMRG step, then perform fDMRG sweepings under the MPS is relaxed to the ground state, and then add 2 more site from the middle, and relax. So the system grows slowly, 2 sites are added each time and then followed by a relax. Hopefully this will converge better. Try this out.
The text was updated successfully, but these errors were encountered:
The current algorithm of iDMRG followed by fDMRG can not easily converge for larger size system. This is because iDMRG is growing too fast, and the MPS generated by iDMRG on the fly is way deviated from the ground state, making the local optimization of fDMRG hard to converge. So we propose a new algorithm here. This is by first growing the system by 2-sites, using an iDMRG step, then perform fDMRG sweepings under the MPS is relaxed to the ground state, and then add 2 more site from the middle, and relax. So the system grows slowly, 2 sites are added each time and then followed by a relax. Hopefully this will converge better. Try this out.
The text was updated successfully, but these errors were encountered: