private string GetOutputCharacter(Pattern p)
{
return solver.GetClosestMatch((BitmapVector)p.Inputs).Information;
}
private BitmapVector GetVector(Pattern p)
{
Bitmap b = ((BitmapVector)p.Inputs).GetBitmap();
DoubleVector dv = DoubleVector.Concatenate(b.GetScaledVectorVerticalHistogram(), b.GetScaledVectorHorizontalHistogram());
return BitmapVector.ConvertToLinearBitmapVector(dv);
}
private string SolveOneCharacter(Pattern p)
{
// Run the pattern through each specialized neural net
sann.ForEach(nn => nn.LastOutput = nn.NeuralNet.FeedForward(p.Inputs.Scale())[0]);
// Sort the networks based on how likely their solution is the solution to the given pattern
sann.Sort((a, b) => -a.LastOutput.CompareTo(b.LastOutput));
// Get the solution of the highest rated network for the given pattern
return sann[0].Solution;
}
private string GetOutputCharacter(Pattern p)
{
return solver.GetClosestMatch(GetVector(p)).Information;
}
/// <summary>
/// Use a winner-takes-all approach to determine what the most likely answer to a pattern is.
/// </summary>
/// <param name="p">The pattern to analyse</param>
/// <returns></returns>
private string GetOutputCharacter(Pattern p)
{
return GetOutputCharacter(p.Outputs);
}
