Building a AI in Autoit

[FMS]

Hello,

I'm building whit mine limited coding know-how a AI in Autoit....
Just because..... why not

It's not doing what i tought it would do.

I hope somebody could help me whit this script?

so far :

-building an learning grid --> AI needs to guess the label in the grin whit only the X and Y value.
-Building an array filled whit random values as weights.
-quess the label
-learn - if quess not the same as the label go change the weights

At this point (the changing of the weights) I've some strange result and hope somebody could point me in the right direction
I think that the problem is in the formula for changing the weights.

PS. I'm also open for good coding practice   I'm learning coding as i go

expandcollapse popup

#AutoIt3Wrapper_Au3Check_Parameters=-d -w 1 -w 2 -w 3 -w 4 -w 5 -w 6 -w 7
#include <MsgBoxConstants.au3>
#include <Array.au3>

build_grid_list()
build_neurons()

_ArrayDisplay($gridlist,"$gridlist 0","",32)
_ArrayDisplay($neurons,"$neurons 0","",32)

For $i = 1 To UBound($gridlist,1) - 1
   If $gridlist[$i][3] = 1 Then
      go_learn($neurons[1][0],$gridlist[$i][0])
   EndIf
Next

_ArrayDisplay($gridlist,"$gridlist 1","",32)
_ArrayDisplay($neurons,"q$neurons 1","",32)

Func go_learn($neuron_ID,$gridlist_ID)
   Local $test = guess($gridlist_ID , $neuron_ID )
   For $i = 1 To 5000
      If $test <> $gridlist[$gridlist_ID][4] Then
         change_wieght($neuron_ID , $gridlist_ID )
         $test = guess($gridlist_ID , $neuron_ID )
      Else
         ExitLoop
      EndIf
   Next
EndFunc

Func guess($gridlist_ID , $neuron_ID = 1 ,$raw = 0);, $loop = 0 )
   Local $temp = 0
;~    For $i = 1 To ubound($neurons,1) -1
      ; input1 * W1 + input2 * W2 + Bias(1) * W3
      $temp = $gridlist[$gridlist_ID][1] * $neurons[$neuron_ID][1] + $gridlist[$gridlist_ID][2] * $neurons[$neuron_ID][2] + 1 * $neurons[$neuron_ID][3]

      ;activate (sign)
      If $temp >= 0 Then
         $neurons[$neuron_ID][6] = 1
      Else
         $neurons[$neuron_ID][6] = -1
      EndIf
      $gridlist[$gridlist_ID][5] = $neurons[$neuron_ID][6]
      $gridlist[$gridlist_ID][6] = $gridlist[$gridlist_ID][6] + 1
      $neurons[$neuron_ID][9] = $neurons[$neuron_ID][9] + 1

      If $raw = 0 Then
         Return $neurons[$neuron_ID][6]
      Else
         Return $temp
      EndIf
EndFunc

Func build_neurons($needed_neurons = 10 )

   Global $neurons[$needed_neurons + 1][10]
   $neurons[0][0] = "id"
   $neurons[0][1] = "wieght1"
   $neurons[0][2] = "wieght2"
   $neurons[0][3] = "wieght3"
   $neurons[0][4] = "input1"
   $neurons[0][5] = "input2"
   $neurons[0][6] = "output"
   $neurons[0][7] = "tweak_counter"
   $neurons[0][8] = "not_tweak_counter"
   $neurons[0][9] = "quess_counter"

   Local $temp = 1
   For $i = 1 To $needed_neurons; -1
      $neurons[$i][0] = $i                  ;id

      Do                                    ;zero_check output wieght1
         $temp = Random(-1 , 1)
      Until $temp <> 0
      $neurons[$i][1] = $temp               ;wieght1

      Do                                    ;zero_check output wieght2
         $temp = Random(-1 , 1)
      Until $temp <> 0
      $neurons[$i][2] = $temp               ;wieght2

      Do                                    ;zero_check output wieght3
         $temp = Random(-1 , 1)
      Until $temp <> 0
      $neurons[$i][3] = $temp               ;wieght3

      $neurons[$i][4] = 0                   ;input1
      $neurons[$i][5] = 0                   ;input2
      Do                                    ;zero_check output
         $temp = Random(-1 , 1 , 1)
      Until $temp <> 0
      $neurons[$i][6] = $temp               ;output +1 / -1
;~    $neurons[$i][6] = 0                   ;output
      $neurons[$i][7] = 0                   ;tweak_counter
      $neurons[$i][8] = 0                   ;not_tweak_counter
      $neurons[$i][9] = 0                   ;quess_counter
   Next
EndFunc

Func build_grid_list($grid_x = 10 ,$grid_y = 10 )
   Global $gridlist[($grid_x * $grid_y) + 1 ][7]
   Local $counter = 1
   $gridlist[0][0] = "ID"
   $gridlist[0][1] = "X"
   $gridlist[0][2] = "Y"
   $gridlist[0][3] = "Active"
   $gridlist[0][4] = "Label"
   $gridlist[0][5] = "quessed"
   $gridlist[0][6] = "quessed_counter"
   For $x = 0 to $grid_x - 1
      For $y = 0 to $grid_y - 1
         $gridlist[$counter][0] = $counter
         $gridlist[$counter][1] = $x
         $gridlist[$counter][2] = $y
         If Random(-1 , 1) >= 0 Then
            $gridlist[$counter][3] = 0
         Else
            $gridlist[$counter][3] = 1
         EndIf
         If $x > $y Then
            $gridlist[$counter][4] = 1
         Else
            $gridlist[$counter][4] = -1
         EndIf
         $gridlist[$counter][5] = -99
         $gridlist[$counter][6] = 0
         $counter = $counter + 1
      Next
   Next
EndFunc

Func change_wieght($neuron_id , $grid_id );, $W1 , $W2 )
   ;W1 = W1 + ^W1 (some change in W1)
   ;^W = err(known) * input ($neurons[$id][3] = "output") * learningrate
   ;$neurons[$id][1] = $neurons[$id][1] + "wieght1"
   Local $iReturn = False

   ;  Desired   |   Quess   |   Error
   ;    -1          -1          0
   ;    -1          +1          -2
   ;    +1          -1          +2
   ;    +1          +1          0
   Local $error = $gridlist[$grid_id][4] - $neurons[$neuron_id][6]
   If $error <> 0 Then
      Local $learningrate = 0.1
      Local $str_len1 = StringLen($neurons[$neuron_id][1])
      Local $str_len2 = StringLen($neurons[$neuron_id][2])
      Local $str_len3 = StringLen($neurons[$neuron_id][3])
      Local $dif_weights1 = $error * $neurons[$neuron_id][1] * $learningrate
      Local $dif_weights2 = $error * $neurons[$neuron_id][2] * $learningrate
      Local $dif_weights3 = $error * $neurons[$neuron_id][3] * $learningrate
      Local $verschil1 = StringLeft($dif_weights1,$str_len1)
      Local $verschil2 = StringLeft($dif_weights2,$str_len2)
      Local $verschil3 = StringLeft($dif_weights3,$str_len3)
      Local $new_wieght1 = $neurons[$neuron_id][1] + $verschil1
      Local $new_wieght2 = $neurons[$neuron_id][2] + $verschil2
      Local $new_wieght3 = $neurons[$neuron_id][3] + $verschil3
      $neurons[$neuron_id][1] = StringLeft($new_wieght1,$str_len1) 
      $neurons[$neuron_id][2] = StringLeft($new_wieght2,$str_len2)
      $neurons[$neuron_id][3] = StringLeft($new_wieght3,$str_len3)

      $neurons[$neuron_id][7] = $neurons[$neuron_id][7] + 1 ;"counter"
      $iReturn = False
   Else
      $neurons[$neuron_id][8] = $neurons[$neuron_id][8] + 1 ;"not counter"
      $iReturn = True
   EndIf

   Return $iReturn

EndFunc

thanks in advanced.

 

[DynamicRookie]

Neural Networks is the most slow and complex algorithm of AI Development (There are also variations that can be more or less complex)  in my opinion.

To calculate weights you must have special algorithms suited for ur purpose and ur database properties (I deduce you are trying to build a Chatbot), i can't really tell what the problem in ur code is, i am not really skilled in those type of areas, however, i'm pretty sure a shorter algorithm for building neurons would work fine. 

 

What makes Neural Networks complex:

Spoiler

each sentence is broken down into different words and each word then is used as input for the neural networks. The weighted connections are then calculated by different iterations through the training data thousands of times. Each time improving the weights to making it accurate. The trained data of neural network is a comparable algorithm more and less code. When there is a big sample, where the training sentences have 20450 different words and 2450 classes, then that would be a matrix of 20450×2450. But this matrix size increases by n (X number) times more gradually and can cause a huge number of errors. In this kind of situations, processing speed should be considerably high, that would cause ur Artificial Intelligence to work slower with the time, that's the reason why i think Neural Networks is the most complex.

 

Let me know if you fixed your problem!