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
When performing large-scale scans I noticed something odd:
In the EndPoint structure, the supreme field sometimes seems to be a complicated AD structure (an example is provided at the end). Is this something that I should expect?
When this happens, it also always seems to hold: value is nothing status is :LOSS_ERROR_STOP
If you are very interested, I could try to provide more information (but a bit hard as I saw this when checking the result of late scans). For now, two questions:
Is this something that I should expect, or something odd?
Can I just interpret this as that LiklihoodProfiler could not finish, and just set the endpoint to the bound?
I should also mention that this happened while doing gradient-based computation of itnervals.
LOSS_ERROR_STOP stands for NLopt FORCED_STOP which is a marker of some error in the objective function or gradient computations. So the result is not reliable and you may try different algorithm/ gradient computation backend, etc.
For the Dual type in the supreme, it seems a missing type check in the code - will fix it
When performing large-scale scans I noticed something odd:
In the
EndPoint
structure, thesupreme
field sometimes seems to be a complicated AD structure (an example is provided at the end). Is this something that I should expect?When this happens, it also always seems to hold:
value
isnothing
status
is:LOSS_ERROR_STOP
If you are very interested, I could try to provide more information (but a bit hard as I saw this when checking the result of late scans). For now, two questions:
I should also mention that this happened while doing gradient-based computation of itnervals.
The text was updated successfully, but these errors were encountered: