Fix unstable results in hist method #4767
Closed
+7
−5
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
After my changes I observed some small differences in results from run to run.
For example, log-loss for HIGGS data set:
1 run: 0.807342768946606615
2 run: 0.807342768708895875
The reason was so easy - guided scheduling instead of static in OpenMP loop. It leads to non-deterministic operations with floating-point numbers.
I have fixed it and now log-loss:
1 run: 0.807342767697801778
2 run: 0.807342767697801778
3 run: 0.807342767697801778
Training time is the same as it were before:
Before time: 2951ms
After time: 2903ms