Hierarchical Multiobjective Pool Solution Runtime
OngoingHow can I retrieve the runtime for each solution found in a biobjective model and not just the time it takes to optimize each respective objective? Observe below for an example of retrieved times of objective optimization through a callback but no information on individual solution run times.
-
Official comment
-
Gurobi Staff
This post is more than two years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support. -
-
-
Hi Brandon,
I guess you are interested at knowing the time it takes to find every feasible solution. When solving a multi-objective optimization model, you still can call the MIP-related callbacks. You can use the MIPSOL callback which is invoked upon finding a new incumbent solution and save the current time as the time it takes to find the newly found incumbent solution.
Best regards,
Maliheh
0 -
-
Hi Maliheh,
I do not think the MIPSOL callback will solve this issue. I'm interested if there is anyway to do something along the lines of "model.params.SolutionNumber = s" for each solution s and retrieve model.Runtime from there at a where == GRB. Callback. MULTIOBJ during the multiobjective solution process. From my understanding model.params.SolutionNumber allows you to retrieve solution vector for the model but not the runtime associated to reach this solution vector.
0 -
-
-
Hi Brandon,
Please see the code snippet below for what I exactly meant. Maybe I am misunderstanding your question. Could you please clarify what else you would like to do which the code below does not help with?
def callback(model, where):
if where == GRB.Callback.MULTIOBJ:
model._obj_count = model.cbGet(GRB.Callback.MULTIOBJ_OBJCNT)
if where == GRB.Callback.MIPSOL:
new_solution = model.cbGetSolution(model._vars)
# Ensure that this is a new solution not a MIPStart for the next objective
if new_solution not in model._solutions:
time = model.cbGet(GRB.Callback.RUNTIME)
model._solutions.append(new_solution)
print(
f"Found a new solution at time {time} while optimizing objective #{model._obj_count + 1}"
)
# Initialize the current count of the objective started to be optimized
model._obj_count = 0
model._vars = model.getVars()
model._solutions = []
model.optimize(callback)Best regards,Maliheh0 -
-
Hi Maliheh,
When I implement your code using both the model._vars or model._specific_decision_variable I cannot retrieve any vectors to confirm the existence of a new solution at GRB.Callback.MIPSOL. I get the following output:
If I use the below pseudo code after the entire model has been optimized I can retrieve solutions and confirm that indeed the solution vector of "s" is different from "s+1." However model.Runtime is equivalent for each solution when you look at the file where metrics are stored. This is supported by the 'OBJ_1_Val, OBJ_2_Val' columns containing decreasing values but 'Sol-Time' staying constant in the second screenshot.
for s in range(model.SolCount):
# Set which solution we will query from now on
model.params.SolutionNumber = s
# generate object with solution vector of solution s
object_s = object_generator([model.getAttr('Xn',model._X).items() for X in set_of_spcific_variables])
with open(results_file, mode='a') as results:
[store object_s metrics to .csv file such as model.Runtime of solution s]
results.close()
Looking at 'In-Acc, 'Out-Acc, Obj_2_Val' it is quite clear how the 10 solutions generated by the model are indeed unique in their solution vectors but I cannot store any runtime information w.r.t the specific solutions. This is supported by the following screenshot showing solution vectors.
0 -
Post is closed for comments.
Comments
5 comments