Question

I want to learn how does an Adaline neuron works.

I written a sample code with VB.NET, and I want to train AND function or OR function for a neuron.

Here is my code :

    Dim Valid As Boolean
    Dim c As Integer = 0
    Do
        Valid = True
        c += 1

        For i As Integer = 0 To ListBox1.Items.Count - 1
            Dim s() As String = Microsoft.VisualBasic.Split(ListBox1.Items(i), " ")
            Dim y As Single = (W(1) * CInt(s(0)) + W(2) * CInt(s(1)))' + W(3) * 1)
            W(1) += Alpha * CSng(CInt(s(2)) - y) * CSng(CInt(s(0)))
            W(2) += Alpha * CSng(CInt(s(2)) - y) * CSng(CInt(s(1)))
            'W(3) += Alpha * CSng(CInt(s(2)) - y) * 1.0
        Next
        For i As Integer = 0 To ListBox1.Items.Count - 1
            Dim s() As String = Microsoft.VisualBasic.Split(ListBox1.Items(i), " ")
            Dim y As Single = (W(1) * CInt(s(0)) + W(2) * CInt(s(1)))' + W(3) * 1)
            If CInt(s(2)) = 1 Then
                If y < 1 Then
                    Valid = False
                End If
            Else
                If y >= 1 Then
                    Valid = False
                End If
            End If
        Next
    Loop Until ((Valid) Or (c > 100000))
    If c > 10000 Then
        MsgBox("I Can't Learn!", MsgBoxStyle.Exclamation)
    End If

I tried with bias and without bias. (If you remove comment mark from the commented parts, a bias will be added to neuron)

and I tried with these training sets for AND & OR:

-1 -1 -1
-1 1 -1
1 -1 -1
1 1 1

0 0 0
0 1 0
1 0 0
1 1 1

0 0 0
0 1 1
1 0 1
1 1 1

-1 -1 -1
-1 1 1
1 -1 1
1 1 1

But the neuron can’t trained with any of these training sets…

How do I solve my problem?

Answer 1

The problem is :
My hard limit function changes the output in 1, but it should change the answer at 0.

This is the changed part of code:

        If CInt(s(2)) = 1 Then
            If y < 0 Then
                Valid = False
            End If
        Else
            If y >= 0 Then
                Valid = False
            End If

I tried it with bias enabled and it successfully trained AND & OR functions.