change the simplex_queue to a SortedKeyList
I suddenly realized the reason why the test_learner_performance_is_invariant_under_scaling was failing for the LearnerND sometimes.
We were saving the rounded loss and then when a subsimplex was created from an existing simplex, we divided the rounded loss instead of dividing the unrounded loss and then rounding it again.
Using a SortedKeyList will make sure this doesn't happen.
Merge request reports
Activity
added 1 commit
- 0d1c1dba - take fewer points to speed up the test for the LearnerND
If were doing it like this, wouldn't it be better to use
sortedcollections.ItemSortedDictjust like in !139 (merged), then we also do not need to keep track of the losses and the queue in a separate place.@Jorn I agree, however that would be a bit more work and I would prefer to merge this as is and implement the
sortedcollections.ItemSortedDictat some later time.100 100 return point 101 101 102 102 103 def _simplex_evaluation_priority(key): 104 # We round the loss to 8 digits such that losses 105 # are equal up to numerical precision will be considered 106 # to be equal. changed this line in version 8 of the diff
100 100 return point 101 101 102 102 103 def _simplex_evaluation_priority(key): 104 # We round the loss to 8 digits such that losses 105 # are equal up to numerical precision will be considered 106 # to be equal. 107 loss, simpl, subsimpl = key changed this line in version 8 of the diff
mentioned in issue #133
enabled an automatic merge when the pipeline for 590d90f5 succeeds
mentioned in commit b9009142
