What's going wrong with this button!

Question

I'm implementing an Adaline algorithm where all inputs are 1x5526 and the output is a number from 1 to 10. I have this two functions defined and worked outside the Manipulate environment:

test[in_, weights_] := (a = Round[Total[Most[weights]*Most[in]] + Last[weights]];  Return[a])
train[in_, out_, alpha_, delta_] := ((*uweights holds also the bias with 1*)      
sizePattern = Length[in[[1]]];
sizeOutput = 1;
Length[out];
uweights = ConstantArray[0, {sizePattern}];
dweights = {};
flag = True;
cont = 0;
Error = {};
While[flag == True && cont < 10000,
    flag = False;
    perror = 0;
    cont++;
    (*Checking each input-output*)
For[i = 1, i <= 1(*Dimensions[out][[1]]*), i++,
    (*Checking each neuron*)
    For[k = 1, k <= Length[out](*Dimensions[out][[2]]*), k++,
        est = Total[Most[uweights]*Most[in[[k]]]] + Last[uweights];
        If[out[[k]] + delta <= est || out[[k]] - delta >= est,
            (*Print["Estimated different"] ;*)
            flag = True;
            dweights = alpha*1*(out[[k]] - est)*Most[in[[k]]];
            dbias = alpha*1*(out[[k]] - est);
            perror = perror + .5*(out[[k]] - est)^2;
            uweights = uweights + Flatten[{dweights, dbias}]]]]];
Return[uweights])

In these lines, they gave the amount of good answer I get with these weights:

AbsoluteTiming[w = train[Samples, AOutput, 0.00001, 0.4];]
uweights = w;
Report = Table[test[Samples[[i]], w], {i, 30}];
Count[Table[Report[[i]] == AOutput[[i]], {i, 30}], True]

{28.516558, Null}

30

The problem emerges when I put the functions test and train inside the Manipulate environment:

Manipulate[
    Grid[{{Text@Style["Saída ", "Large"]}, {Text@Style[answer, "Large"], 
        Text@Style["    ", "Large"]}, 
        {ArrayPlot[ArrayReshape[Samples[[dato]], {85, 65}]]}, 
        (*This line was add to verify if train is working!!!*)
        {weights // MatrixForm}}],
    {{weights, {}}, ControlType -> None},
    {{answer, {}}, ControlType -> None},
    {{dato, 1}, ControlType -> None},
    {{alpha, 0.00001, "Learning rate "}, 0.0000001, .0001, 0.0000005, 
    Appearance -> "Labeled"},
    {{delta, 0.5, "Error range "}, 0.01, 0.5, 0.01, 
    Appearance -> "Labeled"},
    Row[{PopupMenu[Dynamic[NumberPicked], Range[0, 9]], 
       PopupMenu[
        Dynamic[FontType], {1 -> "Comics", 2 -> "Times", 3 -> "Calibri"}]}],
    Row[{Button["Train", weights = train[Samples, AOutput, alpha, delta],
    Background -> Yellow, ImageSize -> Medium],
    Style["                    ", 14],
    Button["Test",
        If[FontType == 1, 
          Switch[NumberPicked, 0, dato = 1, 1, dato = 4, 2, dato = 7, 3, 
          dato = 10, 4, dato = 13, 5, dato = 16, 6, dato = 19, 7, 
          dato = 22, 8, dato = 25, 9, dato = 28]];
        If[FontType == 2, 
          Switch[NumberPicked, 0, dato = 2, 1, dato = 5, 2, dato = 8, 3, 
          dato = 11, 4, dato = 14, 5, dato = 17, 6, dato = 20, 7, 
          dato = 23, 8, dato = 26, 9, dato = 29]];
        If[FontType == 3, 
          Switch[NumberPicked, 0, dato = 3, 1, dato = 6, 2, dato = 9, 3, 
          dato = 12, 4, dato = 15, 5, dato = 18, 6, dato = 21, 7, 
          dato = 24, 8, dato = 27, 9, dato = 30]];
        answer = test[Samples[[dato]], weights],
        Background -> Green, ImageSize -> Medium]}]]

And the output is this:

enter image description here

After push the button train, it seems that the train function was executed, but no values are returned as you can see below the zero. I appreciate any help with this! maybe it seems a silly mistake but I cannot see it.

If the train calculation takes longer than 5 seconds the button will be timing out. You might need to use the Method -> "Queued" option. — Simon Woods, Apr 30 '14 at 20:32
This question appears to be off-topic because it is too localized; i.e, it applies only to the local situation and needs of its poster and answers will not benefit others. — m_goldberg, May 01 '14 at 02:58
m_goldberg it would be nice that you indicate a link toward the duplicated question! — Jotasmall, May 01 '14 at 11:29

score 0 · Answer 1 · answered Apr 30 '14 at 04:10

I suggest that you wrap the definition of the functions into the Module or with if there are many internal variables:

test[in_, weights_] := Round[Total[Most[weights]*Most[in]] + Last[weights]]

train[in_, out_, alpha_, delta_] := 
     Module[{sizePattern, sizeOutput, uweights, dweights, flag, 
cont, Error, perror, i, k, est, dbias},
…
Return[uweights]]

Sometimes the scoping issue will cause the problem. Is there a way to share the data you are using to assist the debugging?

Nope, it is a set of 200 images of 85x65! thanks, but your answer doesn't solve this problem. — Jotasmall, May 01 '14 at 11:16

What's going wrong with this button!

1 Answers1