public unsafe double getRandomAccuracy(IFeedResultNN nn, double level, int sampleSize)
{
int incorrectResults = 0;
//for every training input array
for (int tp = 0; tp < sampleSize; tp++)
{
int index = rand.Next(size);
//feed inputs through network and backpropagate errors
var ot = nn.feedResult(inputs[index]);
//correct pattern flag
//check all outputs against desired output values
for (int k = 0; k < ot.Length; k++)
{
//set flag to false if desired and output differ
if (ot[k] > outputs[index][k] + level || ot[k] < outputs[index][k] - level)
{
incorrectResults++;
break;
}
}
//inc training error for a incorrect result
} //end for
//calculate error and return as percentage
return(1 - incorrectResults / (double)(sampleSize));
}
static void printAnswer(IFeedResultNN nn)
{
double[] first = new double[nI] {
0.1, -0.7, 0.3, 0.99, -0.2, -0.4, 0.3, 0.1, -0.9, -0.2
};
WriteLine(string.Join(" ", nn.feedResult(first).Select(o => o.ToString("N2"))));
}
public unsafe double getMSE(IFeedResultNN nn)
{
double mse = 0;
double[] ot = new double[0];
//for every training input array
for (int tp = 0, l = size; tp < l; tp++)
{
//feed inputs through network and backpropagate errors
ot = nn.feedResult(inputs[tp]);
//check all outputs against desired output values
for (int k = 0; k < ot.Length; k++)
{
//sum all the MSEs together
mse += Abs(ot[k] - outputs[tp][k]);
}
} //end for
//calculate error and return as percentage
return(mse / (ot.Length * size));
}
public neuralPlayer(IFeedResultNN net, StdSerializers.ISerializer ser) : base((arr) => net.feedResult(arr)[0], ser)
{
nn = net;
}
