Skip to content

Adding more metrics to BinaryClassification Experiment #6571

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 4 commits into from
Mar 9, 2023
Merged

Adding more metrics to BinaryClassification Experiment #6571

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8075f39
fix #6570
LittleLittleCloud Feb 16, 2023
f9f075d
Merge branch 'main' into u/xiaoyun/02_16_fix#6570
LittleLittleCloud Feb 22, 2023
1aacc93
fix build error
LittleLittleCloud Mar 7, 2023
0b19772
fix build error
LittleLittleCloud Mar 8, 2023
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
37 changes: 19 additions & 18 deletions src/Microsoft.ML.AutoML/API/BinaryClassificationExperiment.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -381,14 +381,8 @@ public TrialResult Run(TrialSettings settings)
// now we just randomly pick a model, but a better way is to provide option to pick a model which score is the cloest to average or the best.
var res = metrics[_rnd.Next(fold)];
var model = res.Model;
var metric = metricManager.Metric switch
{
BinaryClassificationMetric.PositivePrecision => res.Metrics.PositivePrecision,
BinaryClassificationMetric.Accuracy => res.Metrics.Accuracy,
BinaryClassificationMetric.AreaUnderRocCurve => res.Metrics.AreaUnderRocCurve,
BinaryClassificationMetric.AreaUnderPrecisionRecallCurve => res.Metrics.AreaUnderPrecisionRecallCurve,
_ => throw new NotImplementedException($"{metricManager.MetricName} is not supported!"),
};
var metric = GetMetric(metricManager.Metric, res.Metrics);

var loss = metricManager.IsMaximize ? -metric : metric;
stopWatch.Stop();

Expand All @@ -413,16 +407,7 @@ public TrialResult Run(TrialSettings settings)
var model = pipeline.Fit(trainTestDatasetManager.TrainDataset);
var eval = model.Transform(trainTestDatasetManager.TestDataset);
var metrics = _context.BinaryClassification.EvaluateNonCalibrated(eval, metricManager.LabelColumn, predictedLabelColumnName: metricManager.PredictedColumn);

// now we just randomly pick a model, but a better way is to provide option to pick a model which score is the cloest to average or the best.
var metric = Enum.Parse(typeof(BinaryClassificationMetric), metricManager.MetricName) switch
{
BinaryClassificationMetric.PositivePrecision => metrics.PositivePrecision,
BinaryClassificationMetric.Accuracy => metrics.Accuracy,
BinaryClassificationMetric.AreaUnderRocCurve => metrics.AreaUnderRocCurve,
BinaryClassificationMetric.AreaUnderPrecisionRecallCurve => metrics.AreaUnderPrecisionRecallCurve,
_ => throw new NotImplementedException($"{metricManager.Metric} is not supported!"),
};
var metric = GetMetric(metricManager.Metric, metrics);
var loss = metricManager.IsMaximize ? -metric : metric;

stopWatch.Stop();
Expand Down Expand Up @@ -465,5 +450,21 @@ public Task<TrialResult> RunAsync(TrialSettings settings, CancellationToken ct)
throw;
}
}

private double GetMetric(BinaryClassificationMetric metric, BinaryClassificationMetrics metrics)
{
return metric switch
{
BinaryClassificationMetric.PositivePrecision => metrics.PositivePrecision,
BinaryClassificationMetric.Accuracy => metrics.Accuracy,
BinaryClassificationMetric.AreaUnderRocCurve => metrics.AreaUnderRocCurve,
BinaryClassificationMetric.AreaUnderPrecisionRecallCurve => metrics.AreaUnderPrecisionRecallCurve,
BinaryClassificationMetric.PositiveRecall => metrics.PositiveRecall,
BinaryClassificationMetric.NegativePrecision => metrics.NegativePrecision,
BinaryClassificationMetric.NegativeRecall => metrics.NegativeRecall,
BinaryClassificationMetric.F1Score => metrics.F1Score,
_ => throw new NotImplementedException($"{metric} is not supported!"),
};
}
}
}
34 changes: 15 additions & 19 deletions src/Microsoft.ML.AutoML/API/MulticlassClassificationExperiment.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -375,15 +375,7 @@ public TrialResult Run(TrialSettings settings)
// now we just randomly pick a model, but a better way is to provide option to pick a model which score is the cloest to average or the best.
var res = metrics[_rnd.Next(fold)];
var model = res.Model;
var metric = metricManager.Metric switch
{
MulticlassClassificationMetric.MacroAccuracy => res.Metrics.MacroAccuracy,
MulticlassClassificationMetric.MicroAccuracy => res.Metrics.MicroAccuracy,
MulticlassClassificationMetric.LogLoss => res.Metrics.LogLoss,
MulticlassClassificationMetric.LogLossReduction => res.Metrics.LogLossReduction,
MulticlassClassificationMetric.TopKAccuracy => res.Metrics.TopKAccuracy,
_ => throw new NotImplementedException($"{metricManager.MetricName} is not supported!"),
};
var metric = GetMetric(metricManager.Metric, res.Metrics);
var loss = metricManager.IsMaximize ? -metric : metric;

stopWatch.Stop();
Expand All @@ -409,16 +401,7 @@ public TrialResult Run(TrialSettings settings)
var model = pipeline.Fit(trainTestDatasetManager.TrainDataset);
var eval = model.Transform(trainTestDatasetManager.TestDataset);
var metrics = _context.MulticlassClassification.Evaluate(eval, metricManager.LabelColumn, predictedLabelColumnName: metricManager.PredictedColumn);

var metric = metricManager.Metric switch
{
MulticlassClassificationMetric.MacroAccuracy => metrics.MacroAccuracy,
MulticlassClassificationMetric.MicroAccuracy => metrics.MicroAccuracy,
MulticlassClassificationMetric.LogLoss => metrics.LogLoss,
MulticlassClassificationMetric.LogLossReduction => metrics.LogLossReduction,
MulticlassClassificationMetric.TopKAccuracy => metrics.TopKAccuracy,
_ => throw new NotImplementedException($"{metricManager.Metric} is not supported!"),
};
var metric = GetMetric(metricManager.Metric, metrics);
var loss = metricManager.IsMaximize ? -metric : metric;

stopWatch.Stop();
Expand Down Expand Up @@ -462,6 +445,19 @@ public Task<TrialResult> RunAsync(TrialSettings settings, CancellationToken ct)
}
}

private double GetMetric(MulticlassClassificationMetric metric, MulticlassClassificationMetrics metrics)
{
return metric switch
{
MulticlassClassificationMetric.MacroAccuracy => metrics.MacroAccuracy,
MulticlassClassificationMetric.MicroAccuracy => metrics.MicroAccuracy,
MulticlassClassificationMetric.LogLoss => metrics.LogLoss,
MulticlassClassificationMetric.LogLossReduction => metrics.LogLossReduction,
MulticlassClassificationMetric.TopKAccuracy => metrics.TopKAccuracy,
_ => throw new NotImplementedException($"{metric} is not supported!"),
};
}

public void Dispose()
{
_context.CancelExecution();
Expand Down
35 changes: 16 additions & 19 deletions src/Microsoft.ML.AutoML/API/RegressionExperiment.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -402,14 +402,7 @@ public Task<TrialResult> RunAsync(TrialSettings settings, CancellationToken ct)
// now we just randomly pick a model, but a better way is to provide option to pick a model which score is the cloest to average or the best.
var res = metrics[_rnd.Next(fold)];
var model = res.Model;
var metric = metricManager.Metric switch
{
RegressionMetric.RootMeanSquaredError => res.Metrics.RootMeanSquaredError,
RegressionMetric.RSquared => res.Metrics.RSquared,
RegressionMetric.MeanSquaredError => res.Metrics.MeanSquaredError,
RegressionMetric.MeanAbsoluteError => res.Metrics.MeanAbsoluteError,
_ => throw new NotImplementedException($"{metricManager.MetricName} is not supported!"),
};
var metric = GetMetric(metricManager.Metric, res.Metrics);
var loss = metricManager.IsMaximize ? -metric : metric;

stopWatch.Stop();
Expand All @@ -434,16 +427,8 @@ public Task<TrialResult> RunAsync(TrialSettings settings, CancellationToken ct)
stopWatch.Start();
var model = pipeline.Fit(trainTestDatasetManager.TrainDataset);
var eval = model.Transform(trainTestDatasetManager.TestDataset);
var res = _context.Regression.Evaluate(eval, metricManager.LabelColumn, scoreColumnName: metricManager.ScoreColumn);

var metric = metricManager.Metric switch
{
RegressionMetric.RootMeanSquaredError => res.RootMeanSquaredError,
RegressionMetric.RSquared => res.RSquared,
RegressionMetric.MeanSquaredError => res.MeanSquaredError,
RegressionMetric.MeanAbsoluteError => res.MeanAbsoluteError,
_ => throw new NotImplementedException($"{metricManager.Metric} is not supported!"),
};
var metrics = _context.Regression.Evaluate(eval, metricManager.LabelColumn, scoreColumnName: metricManager.ScoreColumn);
var metric = GetMetric(metricManager.Metric, metrics);
var loss = metricManager.IsMaximize ? -metric : metric;

stopWatch.Stop();
Expand All @@ -453,10 +438,10 @@ public Task<TrialResult> RunAsync(TrialSettings settings, CancellationToken ct)
{
Loss = loss,
Metric = metric,
Metrics = metrics,
Model = model,
TrialSettings = settings,
DurationInMilliseconds = stopWatch.ElapsedMilliseconds,
Metrics = res,
Pipeline = refitPipeline,
} as TrialResult);
}
Expand All @@ -480,5 +465,17 @@ public void Dispose()
_context.CancelExecution();
_context = null;
}

private double GetMetric(RegressionMetric metric, RegressionMetrics metrics)
{
return metric switch
{
RegressionMetric.RootMeanSquaredError => metrics.RootMeanSquaredError,
RegressionMetric.RSquared => metrics.RSquared,
RegressionMetric.MeanSquaredError => metrics.MeanSquaredError,
RegressionMetric.MeanAbsoluteError => metrics.MeanAbsoluteError,
_ => throw new NotImplementedException($"{metric} is not supported!"),
};
}
}
}