Update Feature Contribution Calculation Samples #3241
Conversation
Codecov Report
@@ Coverage Diff @@
## master #3241 +/- ##
==========================================
- Coverage 72.63% 72.62% -0.01%
==========================================
Files 807 805 -2
Lines 145129 145091 -38
Branches 16220 16220
==========================================
- Hits 105415 105376 -39
Misses 35297 35297
- Partials 4417 4418 +1
|
| var transformedData = transformer.Transform(data); | ||
|
|
||
| // Define a linear trainer. | ||
| var linearTrainer = mlContext.Regression.Trainers.Ols(); |
There was a problem hiding this comment.
mlContext.Regression.Trainers.Ols() [](start = 32, length = 35)
Can we not combined it with the pipeline above? Is there any specific reason for doing so? #Resolved
There was a problem hiding this comment.
I'd like to compute the feature contributions on the transformed data, so I separate this out into two steps.
In reply to: 273270793 [](ancestors = 273270793)
| // Convert training data to IDataView. | ||
| var data = mlContext.Data.LoadFromEnumerable(samples); | ||
|
|
||
| // Create a pipeline to concatenate the features into a feature vector and normalize it. |
There was a problem hiding this comment.
This doesn't look a part of FCC. May we start with raw data?
There was a problem hiding this comment.
I wanted to show that FCC works on featurized data, not on the original columns, so I wanted to make it explicit.
| // Define a feature contribution calculator for all the features, and don't normalize the contributions. | ||
| // These are "trivial estimators" and they don't need to fit to the data, so we can feed a subset. | ||
| var simpleScoredDataset = linearModel.Transform(mlContext.Data.TakeRows(transformedData, 1)); | ||
| var linearFeatureContributionCalculator = mlContext.Transforms.CalculateFeatureContribution(linearModel, normalize: false).Fit(simpleScoredDataset); |
There was a problem hiding this comment.
Is it possible to do Fit(null)? Having simpleScoredDataset is a bit confusing.
There was a problem hiding this comment.
It is confusing, but we cannot use null. We need to pass in something with a schema.
| private class Data | ||
| { | ||
| public float Label { get; set; } | ||
| public float Feature1 { get; set; } |
There was a problem hiding this comment.
You have spaces between public fields in the other sample below.
| yield return data; | ||
| } | ||
| } | ||
| private static double Sigmoid(double x) |
There was a problem hiding this comment.
One line function needs to use => 1.0 / (1.0 + Exp(-x)).
* Updating samples for FCC (cherry picked from commit 6f576de)
ghost
locked as resolved and limited conversation to collaborators
This PR cleans up the samples for FCC and creates a new one specifically for calibrated learners.
Fixes #3233