Global AI and Data Science

Hello,

I solve a scheduling problem with the CP optimizer using Docplex and Python. I can call the propagation function without solving, and then I can retrieve the domains. However, I would like to get the domains after solving the problem with a time limit. After I hit the time limit, I would like to retrieve the variable domains that are still feasible for the non-evaluated branching nodes. The domains should be much tighter than the ones I get from just calling the propagation once. Is there some way to get these values?

------------------------------
Maximilian Kolter
------------------------------

No.

Anyway, there is no guarantee that 1-D projections on the variables of the remaining search space  reduce the domains more than propagation, especially if the problem is large and requires a time limit.

You probably want something else, like a set of additional constraints or nogoods that synthesize the explored space. 

Hello,

I solve a scheduling problem with the CP optimizer using Docplex and Python. I can call the propagation function without solving, and then I can retrieve the domains. However, I would like to get the domains after solving the problem with a time limit. After I hit the time limit, I would like to retrieve the variable domains that are still feasible for the non-evaluated branching nodes. The domains should be much tighter than the ones I get from just calling the propagation once. Is there some way to get these values?

------------------------------
Maximilian Kolter
------------------------------

Original Message:
Sent: 11/16/2023 11:59:00 AM
From: Olivier Lhomme
Subject: RE: CP Optimizier Retrieving Variable Domains

No.

Anyway, there is no guarantee that 1-D projections on the variables of the remaining search space  reduce the domains more than propagation, especially if the problem is large and requires a time limit.

You probably want something else, like a set of additional constraints or nogoods that synthesize the explored space. 

------------------------------
Olivier Lhomme

Original Message:
Sent: Tue November 14, 2023 08:53 AM
From: Maximilian Kolter
Subject: CP Optimizier Retrieving Variable Domains

Hello,

I solve a scheduling problem with the CP optimizer using Docplex and Python. I can call the propagation function without solving, and then I can retrieve the domains. However, I would like to get the domains after solving the problem with a time limit. After I hit the time limit, I would like to retrieve the variable domains that are still feasible for the non-evaluated branching nodes. The domains should be much tighter than the ones I get from just calling the propagation once. Is there some way to get these values?

------------------------------
Maximilian Kolter
------------------------------