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
At the moment we have the following strong issue: when we started to work on monitoring features the algorithms of interest (LKH and EAX) logged incumbent values in every single iteration/generation. I.e. if N is the last iteration we had values for every 1 <= i <= N. Monitoring features so far partially rely on this property (in particular plateau features). Saving the incumbent in every single iteration though is quite space consuming and oftentimes there is little or not progress at all thoughout thousands of generations. To cut a way to long story short: the monitoring feature set code needs adaptation 😃
Example: See listing below. We see that slope features are seem to show reasonable values (negative slope since the trajectory is monotonically decreasing). However, all plateaus features take values Inf, NA or NaN.
No description provided.
The text was updated successfully, but these errors were encountered: