Updated Clustering_Iris F# project to v0.7 and new project structure.

Author: mariuszwojcik

samples/fsharp/common/ConsoleHelper.fs

@@ -0,0 +1,212 @@
+namespace Common
+
+module ConsoleHelper =
+    open System
+    open Microsoft.ML
+    open Microsoft.ML.Runtime.Data
+    open Microsoft.ML.Data
+    open Microsoft.ML.Core.Data
+    open Microsoft.ML.Runtime.Api
+    open System.Reflection
+
+    let printPrediction prediction =
+        printfn "*************************************************"
+        printfn "Predicted : %s" prediction
+        printfn "*************************************************"
+
+    let printRegressionPredictionVersusObserved predictionCount observedCount =
+        printfn "-------------------------------------------------"
+        printfn "Predicted : %d" predictionCount
+        printfn "Actual:     %s" observedCount
+        printfn "-------------------------------------------------"
+
+    let printRegressionMetrics name (metrics : RegressionEvaluator.Result) =
+        printfn "*************************************************"
+        printfn "*       Metrics for %s regression model      " name
+        printfn "*------------------------------------------------"
+        printfn "*       LossFn:        %.2f" metrics.LossFn
+        printfn "*       R2 Score:      %.2f" metrics.RSquared
+        printfn "*       Absolute loss: %.2f" metrics.L1
+        printfn "*       Squared loss:  %.2f" metrics.L2
+        printfn "*       RMS loss:      %.2f" metrics.Rms
+        printfn "*************************************************"
+    
+    let printBinaryClassificationMetrics name (metrics : BinaryClassifierEvaluator.Result) =
+        printfn"************************************************************"
+        printfn"*       Metrics for %s binary classification model      " name
+        printfn"*-----------------------------------------------------------"
+        printfn"*       Accuracy: %.2f%%" (metrics.Accuracy * 100.)
+        printfn"*       Auc:      %.2f%%" (metrics.Auc * 100.)
+        printfn"*       F1Score:  %.2f%%" (metrics.F1Score * 100.)
+        printfn"************************************************************"
+
+    let printMultiClassClassificationMetrics name (metrics : MultiClassClassifierEvaluator.Result) =
+        printfn "************************************************************"
+        printfn "*    Metrics for %s multi-class classification model   " name
+        printfn "*-----------------------------------------------------------"
+        printfn "    AccuracyMacro = %.4f, a value between 0 and 1, the closer to 1, the better" metrics.AccuracyMacro
+        printfn "    AccuracyMicro = %.4f, a value between 0 and 1, the closer to 1, the better" metrics.AccuracyMicro
+        printfn "    LogLoss = %.4f, the closer to 0, the better" metrics.LogLoss
+        printfn "    LogLoss for class 1 = %.4f, the closer to 0, the better" metrics.PerClassLogLoss.[0]
+        printfn "    LogLoss for class 2 = %.4f, the closer to 0, the better" metrics.PerClassLogLoss.[1]
+        printfn "    LogLoss for class 3 = %.4f, the closer to 0, the better" metrics.PerClassLogLoss.[2]
+        printfn "************************************************************"
+
+
+    let private calculateStandardDeviation (values : float array) =
+        let average = values |> Array.average
+        let sumOfSquaresOfDifferences = values |> Array.map(fun v -> (v - average) * (v - average)) |> Array.sum
+        let standardDeviation = Math.Sqrt(sumOfSquaresOfDifferences / float (values.Length-1))
+        standardDeviation;
+
+    let calculateConfidenceInterval95 (values : float array) =
+        let confidenceInterval95 = 1.96 * calculateStandardDeviation(values) / Math.Sqrt(float (values.Length-1));
+        confidenceInterval95
+
+    let printMulticlassClassificationFoldsAverageMetrics algorithmName (crossValResults : (MultiClassClassifierEvaluator.Result * ITransformer * IDataView) array) =
+        
+        let metricsInMultipleFolds = crossValResults |> Array.map(fun (metrics, model, scoredTestData) -> metrics)
+
+        let microAccuracyValues  = metricsInMultipleFolds |> Array.map(fun m -> m.AccuracyMicro)
+        let microAccuracyAverage = microAccuracyValues |> Array.average
+        let microAccuraciesStdDeviation = calculateStandardDeviation microAccuracyValues
+        let microAccuraciesConfidenceInterval95 = calculateConfidenceInterval95 microAccuracyValues
+
+        let macroAccuracyValues = metricsInMultipleFolds |> Array.map(fun m -> m.AccuracyMacro)
+        let macroAccuracyAverage = macroAccuracyValues |> Array.average
+        let macroAccuraciesStdDeviation = calculateStandardDeviation macroAccuracyValues
+        let macroAccuraciesConfidenceInterval95 = calculateConfidenceInterval95 macroAccuracyValues
+
+        let logLossValues = metricsInMultipleFolds |> Array.map (fun m -> m.LogLoss)
+        let logLossAverage = logLossValues |> Array.average
+        let logLossStdDeviation = calculateStandardDeviation logLossValues
+        let logLossConfidenceInterval95 = calculateConfidenceInterval95 logLossValues
+
+        let logLossReductionValues = metricsInMultipleFolds |> Array.map (fun m -> m.LogLossReduction)
+        let logLossReductionAverage = logLossReductionValues |> Array.average
+        let logLossReductionStdDeviation = calculateStandardDeviation logLossReductionValues
+        let logLossReductionConfidenceInterval95 = calculateConfidenceInterval95 logLossReductionValues
+
+        printfn "*************************************************************************************************************"
+        printfn "*       Metrics for %s Multi-class Classification model      " algorithmName
+        printfn "*------------------------------------------------------------------------------------------------------------"
+        printfn "*       Average MicroAccuracy:    %.3f  - Standard deviation: (%.3f)  - Confidence Interval 95%%: (%.3f)" microAccuracyAverage microAccuraciesStdDeviation microAccuraciesConfidenceInterval95
+        printfn "*       Average MacroAccuracy:    %.3f  - Standard deviation: (%.3f)  - Confidence Interval 95%%: (%.3f)" macroAccuracyAverage macroAccuraciesStdDeviation macroAccuraciesConfidenceInterval95
+        printfn "*       Average LogLoss:          %.3f  - Standard deviation: (%.3f)  - Confidence Interval 95%%: (%.3f)" logLossAverage logLossStdDeviation logLossConfidenceInterval95
+        printfn "*       Average LogLossReduction: %.3f  - Standard deviation: (%.3f)  - Confidence Interval 95%%: (%.3f)" logLossReductionAverage logLossReductionStdDeviation logLossReductionConfidenceInterval95
+        printfn "*************************************************************************************************************"
+
+    let printClusteringMetrics name (metrics : ClusteringEvaluator.Result) =
+        printfn "*************************************************"
+        printfn "*       Metrics for %s clustering model      " name
+        printfn "*------------------------------------------------"
+        printfn "*       AvgMinScore: %.4f" metrics.AvgMinScore
+        printfn "*       DBI is: %.4f" metrics.Dbi
+        printfn "*************************************************"
+
+    let consoleWriteHeader (lines : string array) =
+        let defaultColor = Console.ForegroundColor
+        Console.ForegroundColor <- ConsoleColor.Yellow
+        printfn " "
+        for line in lines do
+            printfn "%s" line
+        let maxLength = lines |> Array.map(fun x -> x.Length) |> Array.max
+        printfn "%s" (new string('#', maxLength))
+        Console.ForegroundColor <- defaultColor
+
+    let peekDataViewInConsole<'TObservation when 'TObservation : (new : unit -> 'TObservation) and 'TObservation : not struct> (mlContext : MLContext) (dataView : IDataView) (pipeline : IEstimator<ITransformer>) numberOfRows =
+        
+        let msg = sprintf "Peek data in DataView: Showing %d rows with the columns specified by TObservation class" numberOfRows
+        consoleWriteHeader [| msg |]
+
+        //https://github.com/dotnet/machinelearning/blob/master/docs/code/MlNetCookBook.md#how-do-i-look-at-the-intermediate-data
+        let transformer = pipeline.Fit dataView
+        let transformedData = transformer.Transform dataView
+
+        // 'transformedData' is a 'promise' of data, lazy-loading. Let's actually read it.
+        // Convert to an enumerable of user-defined type.
+        let someRows = 
+            transformedData.AsEnumerable<'TObservation>(mlContext, reuseRowObject = false)
+            // Take the specified number of rows
+            |> Seq.take numberOfRows
+            // Convert to List
+            |> Seq.toList
+
+        someRows
+        |> List.iter(fun row ->
+                        
+                        let lineToPrint =
+                            row.GetType().GetFields(BindingFlags.Instance ||| BindingFlags.Static ||| BindingFlags.NonPublic ||| BindingFlags.Public)
+                            |> Array.map(fun field -> sprintf "| %s: %O" field.Name (field.GetValue(row)))
+                            |> Array.fold (+) "Row--> "
+
+                        printfn "%s" lineToPrint
+            )
+    
+        someRows
+
+    let peekVectorColumnDataInConsole (mlContext : MLContext) columnName (dataView : IDataView) (pipeline : IEstimator<ITransformer>) numberOfRows =
+        let msg = sprintf "Peek data in DataView: : Show %d rows with just the '%s' column" numberOfRows columnName
+        consoleWriteHeader [| msg |]
+
+        let transformer = pipeline.Fit dataView
+        let transformedData = transformer.Transform dataView
+
+        // Extract the 'Features' column.
+        let someColumnData = 
+            transformedData.GetColumn<float32[]>(mlContext, columnName)
+            |> Seq.take numberOfRows
+            |> Seq.toList
+
+        // print to console the peeked rows
+        someColumnData
+        |> List.iter(fun row -> 
+            let concatColumn = 
+                row
+                |> Array.map string
+                |> Array.fold (+) " "
+            printfn "%s" concatColumn
+        )
+                        
+        someColumnData;
+
+    let consoleWriterSection (lines : string array) =
+        let defaultColor = Console.ForegroundColor
+        Console.ForegroundColor <- ConsoleColor.Blue
+        printfn " "
+        lines
+        |> Array.iter (printfn "%s")
+
+        let maxLength = lines |> Array.map(fun x -> x.Length) |> Array.max
+        printfn "%s" (new string('-', maxLength))
+        Console.ForegroundColor <- defaultColor
+    
+    let consolePressAnyKey () =
+        let defaultColor = Console.ForegroundColor
+        Console.ForegroundColor <- ConsoleColor.Green
+        printfn " "
+        printfn "Press any key to finish."
+        Console.ForegroundColor <- defaultColor
+        Console.ReadKey() |> ignore
+    
+    let consoleWriteException (lines : string array) =
+        let defaultColor = Console.ForegroundColor
+        Console.ForegroundColor <- ConsoleColor.Red
+        let exceptionTitle = "EXCEPTION"
+        printfn " "
+        printfn "%s" exceptionTitle
+        printfn "%s" (new string('#', exceptionTitle.Length))
+        Console.ForegroundColor <- defaultColor
+        lines
+        |> Array.iter (printfn "%s")
+
+    let consoleWriteWarning (lines : string array) =
+        let defaultColor = Console.ForegroundColor
+        Console.ForegroundColor <-  ConsoleColor.DarkMagenta
+        let warningTitle = "WARNING"
+        printfn " "
+        printfn "%s" warningTitle
+        printfn "%s" (new string('#', warningTitle.Length))
+        Console.ForegroundColor <- defaultColor
+        lines
+        |> Array.iter (printfn "%s")

samples/fsharp/common/ModelBuilder.fs

@@ -0,0 +1,46 @@
+namespace Common
+
+open System.IO
+open Microsoft.ML
+open Microsoft.ML.Core.Data
+open Microsoft.ML.Transforms
+open Microsoft.ML.Runtime.Data
+
+module ModelBuilder =
+
+    let create (mlContext : MLContext) (pipeline : IEstimator<ITransformer>) =
+        (mlContext, pipeline)
+        
+    let append (estimator : IEstimator<'a>) (pipeline : IEstimator<'b>)  = 
+        match pipeline with
+        | :? IEstimator<ITransformer> as p -> 
+            p.Append estimator
+        | _ -> failwith "The pipeline has to be an instance of IEstimator<ITransformer>."
+
+
+    let addTrainer (trainer : IEstimator<'b>) (mlContext : MLContext, pipeline : IEstimator<'a>) =
+        let newPipeline =
+            pipeline 
+            |> append trainer
+        (mlContext, newPipeline)
+
+    let train (trainingData : IDataView) (mlContext : MLContext, pipeline : IEstimator<'a>) =
+        pipeline.Fit trainingData :> ITransformer
+        
+
+    let private checkTrained (trainedModel : ITransformer) =
+        if (trainedModel = null) then
+            failwith "Cannot test before training. Call Train() first."
+
+    let evaluateClusteringModel (dataView : IDataView) (trainedModel : ITransformer) (mlContext : MLContext, pipeline : IEstimator<'a>) =
+        checkTrained trainedModel
+        let predictions = trainedModel.Transform dataView
+        mlContext.Clustering.Evaluate(predictions, score = "Score", features = "Features")
+
+    let saveModelAsFile persistedModelPath (trainedModel : ITransformer, (mlContext : MLContext, _)) =
+        checkTrained trainedModel
+
+        use fs = new FileStream(persistedModelPath, FileMode.Create, FileAccess.Write, FileShare.Write)
+        mlContext.Model.Save(trainedModel, fs);
+        printfn "The model is saved to %s" persistedModelPath
+

samples/fsharp/common/ModelScorer.fs

@@ -0,0 +1,28 @@
+namespace Common
+
+open System.IO
+open Microsoft.ML
+open Microsoft.ML.Core.Data
+open Microsoft.ML.Transforms
+open Microsoft.ML.Runtime.Data
+
+module ModelScorer =
+
+    let create (mlContext : MLContext) =
+        mlContext
+
+    let loadModelFromZipFile<'TObservation, 'TPrediction when 'TPrediction : (new : unit -> 'TPrediction) and 'TPrediction : not struct and 'TObservation : not struct> modelPath (mlContext : MLContext) =
+        use stream = new FileStream(modelPath, FileMode.Open, FileAccess.Read, FileShare.Read)
+        let trainedModel = TransformerChain.LoadFrom(mlContext, stream)
+        let predictionFunction = trainedModel.MakePredictionFunction<'TObservation, 'TPrediction>(mlContext);
+       
+        mlContext, trainedModel, predictionFunction
+
+    let private checkTrainedModelIsLoaded (trainedModel) =
+        if trainedModel = null then
+            failwith "Need to have a model before scoring. Call LoadModelFromZipFile(modelPath) first or provided a model through the constructor."
+
+    let predictSingle (input : 'TObservation) (mlContext : MLContext, trainedModel, predictionFunction : PredictionFunction<'TObservation, 'TPrediction>) =
+        checkTrainedModelIsLoaded trainedModel
+        predictionFunction.Predict(input);
+        

samples/fsharp/common/Pipeline.fs

@@ -0,0 +1,36 @@
+namespace Common
+
+module Pipeline =
+
+    open Microsoft.ML.Core.Data
+    open Microsoft.ML.Transforms
+    open Microsoft.ML.Runtime.Data
+
+    //let textTransform (inputColumn : string) outputColumn env =
+    //    Microsoft.ML.Transforms.Text.TextTransform(env, inputColumn, outputColumn)
+
+    let copyColumnsEstimator input output env =
+        CopyColumnsEstimator(env, input, output)
+
+    //let concatEstimator input output env =
+    //    ConcatEstimator(env, input, output)
+
+    //let append' (estimator : IEstimator<'b>) (pipeline : IEstimator<ITransformer>)  = 
+    //    pipeline.Append estimator
+
+    let append (estimator : IEstimator<'a>) (pipeline : IEstimator<'b>)  = 
+        match pipeline with
+        | :? IEstimator<ITransformer> as p -> 
+            p.Append estimator
+        | _ -> failwith "The pipeline has to be an instance of IEstimator<ITransformer>."
+
+
+    let fit (dataView : IDataView) (pipeline : EstimatorChain<'a>) =
+        pipeline.Fit dataView
+
+
+    
+    let downcast' (b : IEstimator<'a>) =
+        match b with
+        | :? IEstimator<ITransformer> as b -> b
+        | _ -> failwith "qwe"

samples/fsharp/getting-started/Clustering_Iris/Clustering_Iris.fsproj

@@ -0,0 +1,25 @@
+
+Microsoft Visual Studio Solution File, Format Version 12.00
+# Visual Studio 15
+VisualStudioVersion = 15.0.28010.2050
+MinimumVisualStudioVersion = 10.0.40219.1
+Project("{6EC3EE1D-3C4E-46DD-8F32-0CC8E7565705}") = "Clustering_Iris", "IrisClustering\IrisClusteringConsoleApp\Clustering_Iris.fsproj", "{20A1A0DB-A515-4EF9-891B-C57A4D66AC49}"
+EndProject
+Global
+	GlobalSection(SolutionConfigurationPlatforms) = preSolution
+		Debug|Any CPU = Debug|Any CPU
+		Release|Any CPU = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(ProjectConfigurationPlatforms) = postSolution
+		{20A1A0DB-A515-4EF9-891B-C57A4D66AC49}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
+		{20A1A0DB-A515-4EF9-891B-C57A4D66AC49}.Debug|Any CPU.Build.0 = Debug|Any CPU
+		{20A1A0DB-A515-4EF9-891B-C57A4D66AC49}.Release|Any CPU.ActiveCfg = Release|Any CPU
+		{20A1A0DB-A515-4EF9-891B-C57A4D66AC49}.Release|Any CPU.Build.0 = Release|Any CPU
+	EndGlobalSection
+	GlobalSection(SolutionProperties) = preSolution
+		HideSolutionNode = FALSE
+	EndGlobalSection
+	GlobalSection(ExtensibilityGlobals) = postSolution
+		SolutionGuid = {08E1E2D6-0946-4172-A8D4-1C8B3F691DCD}
+	EndGlobalSection
+EndGlobal