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
Describe the bug
The AcceptanceTracker class, which can be used to track and log the neighbor acceptance rate during runs of simulated annealing assumes that all runs went to end of run. But the SA implementation will terminate early if it finds a solution with cost equal to the theoretical min for the problem. This leads to the potential for acceptance rates to be underreported, especially near end of long runs if some runs terminated early. For example, if there are 100 runs, it will divide by 100 when computing the acceptance rate at iteration i, regardless of whether there were 100 runs that reached iteration i.
Expected behavior
Correct acceptance rate calculation (i.e., divide by number of runs that reached iteration i).
The text was updated successfully, but these errors were encountered:
Describe the bug
The AcceptanceTracker class, which can be used to track and log the neighbor acceptance rate during runs of simulated annealing assumes that all runs went to end of run. But the SA implementation will terminate early if it finds a solution with cost equal to the theoretical min for the problem. This leads to the potential for acceptance rates to be underreported, especially near end of long runs if some runs terminated early. For example, if there are 100 runs, it will divide by 100 when computing the acceptance rate at iteration i, regardless of whether there were 100 runs that reached iteration i.
Expected behavior
Correct acceptance rate calculation (i.e., divide by number of runs that reached iteration i).
The text was updated successfully, but these errors were encountered: