Skip to content

Fix PFI issue in binary classification #4587

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 10 commits into from
Jan 8, 2020
Merged

Fix PFI issue in binary classification #4587

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
4ed3ca8
Add non-calibrated evaluation to PFI
yaeldMS Dec 18, 2019
fe48872
change to always call EvaluateNonCalibrated
yaeldMS Jan 2, 2020
b6ed4b3
Add non-calibrated evaluation to PFI
yaeldMS Dec 18, 2019
0559fb7
change to always call EvaluateNonCalibrated
yaeldMS Jan 2, 2020
6558551
Add non-calibrated evaluation to PFI
yaeldMS Dec 18, 2019
2d53161
change to always call EvaluateNonCalibrated
yaeldMS Jan 2, 2020
bd2af8e
Add non-calibrated evaluation to PFI
yaeldMS Dec 18, 2019
6e53e5b
Add asserts to unit test
yaeldMS Jan 5, 2020
8bde46b
Remove using statements
yaeldMS Jan 5, 2020
cd7f27c
Remove unused ctor
yaeldMS Jan 7, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion src/Microsoft.ML.Transforms/PermutationFeatureImportance.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -171,7 +171,7 @@ public static ImmutableArray<TResult>
int processedCnt = 0;
int nextFeatureIndex = 0;
var shuffleRand = RandomUtils.Create(host.Rand.Next());
using (var pch = host.StartProgressChannel("SDCA preprocessing with lookup"))
using (var pch = host.StartProgressChannel("Calculating Permutation Feature Importance"))
{
pch.SetHeader(new ProgressHeader("processed slots"), e => e.SetProgress(0, processedCnt));
foreach (var workingIndx in workingFeatureIndices)
Expand Down
20 changes: 10 additions & 10 deletions src/Microsoft.ML.Transforms/PermutationFeatureImportanceExtensions.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,16 +145,16 @@ public static ImmutableArray<BinaryClassificationMetricsStatistics>
int permutationCount = 1) where TModel : class
{
return PermutationFeatureImportance<TModel, BinaryClassificationMetrics, BinaryClassificationMetricsStatistics>.GetImportanceMetricsMatrix(
catalog.GetEnvironment(),
predictionTransformer,
data,
() => new BinaryClassificationMetricsStatistics(),
idv => catalog.Evaluate(idv, labelColumnName),
BinaryClassifierDelta,
predictionTransformer.FeatureColumnName,
permutationCount,
useFeatureWeightFilter,
numberOfExamplesToUse);
catalog.GetEnvironment(),
predictionTransformer,
data,
() => new BinaryClassificationMetricsStatistics(),
idv => catalog.EvaluateNonCalibrated(idv, labelColumnName),
Copy link
Contributor

@harishsk harishsk Jan 3, 2020 •
edited by yaeldekel
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Small Nit:
The only difference between the if (isCalibratedModel) and the else case is the idv parameter. Is it possible to make this a bit more readable by factoring out just that line and using a single call to the PermutationFeatureImportance constructor? #Resolved

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Resolving this comment, since isCalibratedModel has been removed.

In reply to: 362685943 [](ancestors = 362685943)

BinaryClassifierDelta,
predictionTransformer.FeatureColumnName,
permutationCount,
useFeatureWeightFilter,
numberOfExamplesToUse);
}

private static BinaryClassificationMetrics BinaryClassifierDelta(
Expand Down
30 changes: 30 additions & 0 deletions test/Microsoft.ML.Tests/PermutationFeatureImportanceTests.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -305,6 +305,36 @@ public void TestPfiBinaryClassificationOnSparseFeatures(bool saveModel)

Done();
}

[Fact]
public void TestBinaryClassificationWithoutCalibrator()
Copy link
Contributor

@harishsk harishsk Jan 3, 2020 •
edited by yaeldekel
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The test does not Assert anything. Can you please include Asserts for the relevant results that this test is supposed to verify? #Resolved

{
var dataPath = GetDataPath("breast-cancer.txt");
var ff = ML.BinaryClassification.Trainers.FastForest();
var data = ML.Data.LoadFromTextFile(dataPath,
new[] { new TextLoader.Column("Label", DataKind.Boolean, 0),
new TextLoader.Column("Features", DataKind.Single, 1, 9) });
var model = ff.Fit(data);
var pfi = ML.BinaryClassification.PermutationFeatureImportance(model, data);

// For the following metrics higher is better, so minimum delta means more important feature, and vice versa
Assert.Equal(7, MaxDeltaIndex(pfi, m => m.AreaUnderRocCurve.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.AreaUnderRocCurve.Mean));
Assert.Equal(3, MaxDeltaIndex(pfi, m => m.Accuracy.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.Accuracy.Mean));
Assert.Equal(3, MaxDeltaIndex(pfi, m => m.PositivePrecision.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.PositivePrecision.Mean));
Assert.Equal(3, MaxDeltaIndex(pfi, m => m.PositiveRecall.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.PositiveRecall.Mean));
Assert.Equal(3, MaxDeltaIndex(pfi, m => m.NegativePrecision.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.NegativePrecision.Mean));
Assert.Equal(2, MaxDeltaIndex(pfi, m => m.NegativeRecall.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.NegativeRecall.Mean));
Assert.Equal(3, MaxDeltaIndex(pfi, m => m.F1Score.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.F1Score.Mean));
Assert.Equal(7, MaxDeltaIndex(pfi, m => m.AreaUnderPrecisionRecallCurve.Mean));
Assert.Equal(1, MinDeltaIndex(pfi, m => m.AreaUnderPrecisionRecallCurve.Mean));
}
#endregion

#region Multiclass Classification Tests
Expand Down