infostreams / neural-network Goto Github PK

Why $n->initWeights();
Every Epoch reset weights ??? Why ??
Reset weights cause Network initialize from new.

And loss previous worked !!

I can't understand if the input values must be between 0 and 1 or between -1 and +1 or... ? can you help me?

protected function activation($value) {
    return tanh($value);
    // return (1.0 / (1.0 + exp(- $value)));
}
protected function derivativeActivation($value) {
    $tanh = tanh($value);
    return 1.0 - $tanh * $tanh;
    //return $value * (1.0 - $value);
}

I don't get it. Either
derivativeActivation should return 1.0 - $value * $value;
or the comment below for sigmoid should read
//return activation($value) * (1.0 - activation($value));

I am not sure understanding the load portion, after saving and loading how to compute? Also how to get output of test data then?

I am not sure if this is an issue or not but..

The following code:
$n->addTestData(array (-1, -1, 1), array (-1));
$n->addTestData(array (-1, 1, 1), array ( 1));
$n->addTestData(array ( 1, -1, 1), array ( 1));
$n->addTestData(array ( 1, 1, 1), array (-1));

Should this not be:
$n->addTestData(array (-1, -1, 0), array (-1));
$n->addTestData(array (-1, 1, 1), array ( 1));
$n->addTestData(array ( 1, -1, 1), array ( 1));
$n->addTestData(array ( 1, 1, 0), array (-1));

If not, what is the purpose of the third variable here if not the classification? If this third variable is not the classification (expected result) then how do you train this NN?

Thanks.