Predicting data in this instance means predicting the argmax value. To do this, we assume that the last layer of the network is a SoftmaxLayer. Prediction occurs when we call the GetPrediction() function, as follows:

public int[] GetPrediction()
{
var softmaxLayer = this._lastLayer as SoftmaxLayer<T>;
if (softmaxLayer == null)
{
throw new Exception("Function assumes softmax as last layer of the net!");
}
var activation = softmaxLayer.OutputActivation;
var N = activation.Shape.Dimensions[3];
var C = activation.Shape.Dimensions[2];
var result = new int[N];
for (varn = 0; n < N; n++)
{
varmaxv = activation.Get(0, 0, 0, n);
varmaxi = 0;
for (vari = 1; i < C; i++)
{
var output = activation.Get(0, 0, i, n);
if (Ops<T>.GreaterThan(output, maxv))
{
maxv = output;
maxi = i;
}
}
result[n] = maxi;
}
return result;
}