/// <summary>
/// This function trains the network that the correct output when given <paramref name="exampleIn"/> should be <paramref name="exampleOut"/>.
/// </summary>
/// <param name="exampleIn">The input of example.</param>
/// <param name="exampleOut">The output of example.</param>
/// <exception cref="ArgumentException">The specified type of kernel is not supported.</exception>
/// <exception cref="ArgumentNullException"><paramref name="exampleIn"/> or <paramref name="exampleOut"/>is null.</exception>
/// <exception cref="ArgumentOutOfRangeException"><paramref name="exampleOut"/> must be 0.0 - 1.0.</exception>
/// <exception cref="ObjectDisposedException"><paramref name="exampleIn"/> or <paramref name="exampleOut"/> is disposed.</exception>
public void Train(Matrix <double> exampleIn, Matrix <double> exampleOut)
{
if (exampleIn == null)
{
throw new ArgumentNullException(nameof(exampleIn));
}
if (exampleOut == null)
{
throw new ArgumentNullException(nameof(exampleOut));
}
exampleIn.ThrowIfDisposed();
exampleOut.ThrowIfDisposed();
var max = Dlib.Max(exampleOut);
var min = Dlib.Min(exampleOut);
if (!(0 <= min && max <= 1.0))
{
throw new ArgumentOutOfRangeException(nameof(exampleOut), $"{nameof(exampleOut)} must be 0.0 - 1.0.");
}
var kernelType = this._MultilayerPerceptronKernelType.ToNativeMlpKernelType();
var ret = NativeMethods.mlp_kernel_train_matrix(kernelType, this.NativePtr, exampleIn.NativePtr, exampleOut.NativePtr);
switch (ret)
{
case NativeMethods.ErrorType.MlpKernelNotSupport:
throw new ArgumentException($"{kernelType} is not supported.");
}
}