Add recall #2518
Add recall #2518
Conversation
…h RecallMetric. Adapt confusionStats and Precision with generalized code.
… rename test function properly in precision.
…ation decision rule and class_reduction
# Conflicts: # crates/burn-train/src/learner/classification.rs # crates/burn-train/src/metric/confusion_stats.rs
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #2518 +/- ##
=======================================
Coverage 81.83% 81.84%
=======================================
Files 836 838 +2
Lines 107214 107302 +88
=======================================
+ Hits 87743 87821 +78
- Misses 19471 19481 +10 ☔ View full report in Codecov by Sentry. |
| /// The prediction decision rule for classification metrics. | ||
| pub enum DecisionRule { | ||
| Threshold(f64), | ||
| TopK(NonZeroUsize), | ||
| } |
There was a problem hiding this comment.
Ok! I think I've come around to the terminology behind this representation.
Just make sure to document both decision rules. Suggestions:
Top-k: Consider the prediction correct if the true label is within the top k predicted classes based on scores.
Threshold: Consider a class predicted if its probability exceeds the threshold.
There was a problem hiding this comment.
Great, I'm glad! Thanks for the tip and suggestions. I changed TopK's slightly. Let me know what you think.
Pull Request Template
Checklist
run-checks allscript has been executed.Related Issues/PRs
#544
Changes
Create
ClassificationInputandClassificationConfig, generalizations ofPrecisionInputandPrecisionConfigfor any classification task (currently used by precision and recall).Add
RecallMetric.Moved ClassReduction into ClassificationConfig, since i think they belong together.
Testing
Added tests for recall