#Region Project Attributes
#MainFormWidth: 600
#MainFormHeight: 600
#End Region
#AdditionalJar: deeplearning4j-core-1.0.0-M2.1.jar
#AdditionalJar: nd4j-native-platform-1.0.0-M2.1.jar
#AdditionalJar: slf4j-api-2.1.0-alpha1.jar
#AdditionalJar: slf4j-simple-2.1.0-alpha1.jar
Sub Process_Globals
Private fx As JFX
Private MainForm As Form
End Sub
Sub AppStart (Form1 As Form, Args() As String)
MainForm = Form1
MainForm.Show
' Costruzione della configurazione della rete neurale
Dim nnConf As JavaObject
nnConf = InitializeNeuralNetConfiguration
If nnConf.IsInitialized Then
Log("Neural network configuration initialized successfully")
Else
Log("Failed to initialize neural network configuration")
End If
' Creazione del modello
Dim model As JavaObject
model.InitializeNewInstance("org.deeplearning4j.nn.multilayer.MultiLayerNetwork", Array(nnConf))
model.RunMethod("init", Null)
' Stampa del sommario del modello
Log(model.RunMethod("summary", Null))
' Esempio di addestramento con dati fittizi
TrainModel(model)
End Sub
Sub InitializeNeuralNetConfiguration As JavaObject
Dim nnConf As JavaObject
nnConf.InitializeNewInstance("org.deeplearning4j.nn.conf.NeuralNetConfiguration$Builder", Null)
nnConf = nnConf.RunMethod("seed", Array(12345)) ' Seed per la randomizzazione dei pesi
nnConf = nnConf.RunMethod("updater", Array(CreateUpdater("adam"))) ' Ottimizzatore Adam
nnConf = nnConf.RunMethod("list", Null)
nnConf = nnConf.RunMethod("layer", Array(0, CreateDenseLayer(8, 26, "relu"))) ' Primo layer
nnConf = nnConf.RunMethod("layer", Array(1, CreateDenseLayer(26, 26, "relu"))) ' Secondo layer
nnConf = nnConf.RunMethod("layer", Array(2, CreateDenseLayer(26, 26, "relu"))) ' Terzo layer
nnConf = nnConf.RunMethod("layer", Array(3, CreateDenseLayer(26, 8, "relu"))) ' Quarto layer
nnConf = nnConf.RunMethod("layer", Array(4, CreateOutputLayer(8, 8, "softmax"))) ' Output layer
nnConf = nnConf.RunMethod("build", Null)
Return nnConf
End Sub
Sub CreateDenseLayer(numInput As Int, numOutput As Int, activation As String) As JavaObject
Dim denseLayerBuilder As JavaObject
denseLayerBuilder.InitializeNewInstance("org.deeplearning4j.nn.conf.layers.DenseLayer$Builder", Null)
denseLayerBuilder = denseLayerBuilder.RunMethod("nIn", Array(numInput))
denseLayerBuilder = denseLayerBuilder.RunMethod("nOut", Array(numOutput))
denseLayerBuilder = denseLayerBuilder.RunMethod("activation", Array(activation))
denseLayerBuilder = denseLayerBuilder.RunMethod("weightInit", Array("xavier"))
Return denseLayerBuilder.RunMethod("build", Null)
End Sub
Sub CreateOutputLayer(numInput As Int, numOutput As Int, activation As String) As JavaObject
Dim outputLayerBuilder As JavaObject
outputLayerBuilder.InitializeNewInstance("org.deeplearning4j.nn.conf.layers.OutputLayer$Builder", Null)
outputLayerBuilder = outputLayerBuilder.RunMethod("nIn", Array(numInput))
outputLayerBuilder = outputLayerBuilder.RunMethod("nOut", Array(numOutput))
outputLayerBuilder = outputLayerBuilder.RunMethod("activation", Array(activation))
outputLayerBuilder = outputLayerBuilder.RunMethod("weightInit", Array("xavier"))
outputLayerBuilder = outputLayerBuilder.RunMethod("lossFunction", Array("negativeloglikelihood"))
Return outputLayerBuilder.RunMethod("build", Null)
End Sub
Sub CreateUpdater(updaterType As String) As JavaObject
Dim updater As JavaObject
updater.InitializeStatic("org.nd4j.linalg.learning.config." & updaterType.SubString2(0,1).ToUpperCase & updaterType.SubString(1).ToLowerCase)
Return updater
End Sub
Sub TrainModel(model As JavaObject)
' Esempio di dati fittizi per l'addestramento
Dim features(,) As Double = Array(Array(1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0), Array(0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0))
Dim labels(,) As Double = Array(Array(1.0, 0.0), Array(0.0, 1.0))
' Creazione delle matrici di input e output
Dim Nd4j As JavaObject
Nd4j.InitializeStatic("org.nd4j.linalg.factory.Nd4j")
Dim input As JavaObject = Nd4j.RunMethod("create", Array(features))
Dim output As JavaObject = Nd4j.RunMethod("create", Array(labels))
' Addestramento del modello
For i = 1 To 1000
model.RunMethod("fit", Array(input, output))
Next
' Predizione con il modello addestrato
Dim prediction As JavaObject = model.RunMethod("output", Array(input))
Log("Predictions: " & prediction)
End Sub
Sub MainForm_CloseRequest (EventData As Event)
MainForm.Close
End Sub
