public void Regression()
{
ResetClassification();
var watch = System.Diagnostics.Stopwatch.StartNew();
// Create the perceptron
Perceptron model = new Perceptron(2, PerceptronType.REGRESSION);
// Create the training values
double[,] values = new double[reds.Length + blues.Length, 2];
double[] expectedValues = new double[reds.Length + blues.Length];
for (int i = 0; i < reds.Length; i++)
{
Transform t = reds[i].GetComponent <Transform>();
values[i, 0] = t.position.x;
values[i, 1] = t.position.z;
expectedValues[i] = 1.0; // #ff0000
}
for (int i = 0; i < blues.Length; i++)
{
Transform t = blues[i].GetComponent <Transform>();
values[i + reds.Length, 0] = t.position.x;
values[i + reds.Length, 1] = t.position.z;
expectedValues[i + reds.Length] = -1.0; // #0000ff
}
model.Train(0.001, values, expectedValues, 500000);
// Texture generation
Transform tRenderer = renderer.GetComponent <Transform>();
double xWidth = tRenderer.localScale.x;
double yWidth = tRenderer.localScale.z;
double xDec = tRenderer.position.x;
double yDec = tRenderer.position.y;
var texture = new Texture2D(textureSize, textureSize, TextureFormat.ARGB32, false);
for (int x = 0; x < textureSize; x++)
{
for (int y = 0; y < textureSize; y++)
{
double xValue = ((((double)x / textureSize) * xWidth) - (xWidth / 2)) + xDec;
double yValue = ((((double)y / textureSize) * yWidth) - (yWidth / 2)) + yDec;
double val = model.Classify(new double[] { xValue, yValue });
texture.SetPixel(-x, -y, new Color((float)val, 0f, (float)(1.0 - val)));
}
}
texture.Apply();
renderer.GetComponent <Renderer>().material.mainTexture = texture;
watch.Stop();
var elapsedMs = watch.ElapsedMilliseconds;
timerText.text = "" + elapsedMs + " ms";
}