Algorithm can exit prematurely when optim() causes an individual to violate a constraint
Bug #184691 reported by
Ben Goodrich
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| FAiR |
Fix Committed
|
High
|
Ben Goodrich | ||
Bug Description
This bug is hard to trigger, but occurs because the gr_fitS4 methods return a gradient of .Machine$double.eps when one of the constraints is violated. It is somehow possible for an individual to be at a location in the parameter space where the constraint holds exactly, move a tiny bit in the wrong direction, stop due to the zero gradient, and nevertheless have the fitS4 be superior to all other individuals for wait.generations.
| Changed in fair: | |
| assignee: | nobody → bgokgm |
| importance: | Undecided → High |
| status: | New → Confirmed |
Revision history for this message
| Ben Goodrich (bgokgm) wrote | #1 |
| Changed in fair: | |
| status: | Confirmed → Fix Committed |
To post a comment you must log in.
This bug was fixed at some point by removing an unnecessary line that was overwriting the correct value.