public void LearnNeverTest()
{
int[] layers = new int[] { 2, 1 }; // no linear separity
MLPWatchdog nn = new WatchdogGenerator().Create(layers, 1, new Sigmoid());
nn.LearnFactor = 0.9;
nn.Momentum = 0.5;
var response = nn.BPWatched(new BPWDRequest(XOR, 10000, 0, 0.1, 10));
Assert.AreEqual(response.Count, 10);
Assert.AreEqual(response[9].Estimate, BPResponseEstimate.Failure);
}
public void LearnXORTest()
{
for (var x = 0; x < 5; x++)
{
int[] layers = new int[] { 2, 2, 1 }; // linear separity
MLPWatchdog nn = new WatchdogGenerator().Create(layers, 1, new Sigmoid());
nn.LearnFactor = 0.9;
nn.Momentum = 0.5;
var response = nn.BPWatched(new BPWDRequest(XOR, 10000, 0, 0.001, 10));
Assert.IsTrue(response.Count < 10); // should learn faster
Assert.AreEqual(response.Last().Estimate, BPResponseEstimate.Success);
//sanity check
IsBetween(nn.Pulse(new double[] { 0, 0 })[0], 0, 0.05);
IsBetween(nn.Pulse(new double[] { 0, 1 })[0], 0.95, 1);
IsBetween(nn.Pulse(new double[] { 1, 0 })[0], 0.95, 1);
IsBetween(nn.Pulse(new double[] { 1, 1 })[0], 0, 0.05);
}
}
