Somewhat strange behavior detected

Hello,

Could please someone shed some light on the following. The code below gives a different result when the line with the arrow is uncommented (error and weights are different), but I'm not sure why.

The goal is to continue a previously interrupted training and not to have the ITrain object in memory.

Thanks!

Regards,
Roger

Here is the code:

BasicNetwork network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationSigmoid(), false, 1));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 1));
network.AddLayer(new BasicLayer(new ActivationSigmoid(), true, 1));
network.Logic = new FeedforwardLogic();
network.Structure.FinalizeStructure();
NetworkCODEC.ArrayToNetwork(new double[] { 0.3233232, 0.1212328, -0.11, 0.984 }, network);
INeuralDataSet trainingSet = new BasicNeuralDataSet(new double[][] { new double[] { 0 } }, new double[][] { new double[] { 0 } });
ITrain train = new Backpropagation(network, trainingSet, 0.3d, 0.9d);
Console.WriteLine("Start at " + DateTime.Now);
double mse = 0;
/********************/
for (int i = 0; i < 1; i++)
{
train.Iteration();
mse = train.Error;
}
//train = new Backpropagation(network, trainingSet, 0.3d, 0.9d); //<---- uncomment this gives different result
for (int i = 0; i < 2; i++)
{
train.Iteration();
mse = train.Error;
}
/********************/
Console.WriteLine("End at " + DateTime.Now);
Console.WriteLine("MSE before last iteration: " + mse);

double[] doubleArray = NetworkCODEC.NetworkToArray(network);
Console.WriteLine(doubleArray[0]);
Console.WriteLine(doubleArray[1]);
Console.WriteLine(doubleArray[2]);
Console.WriteLine(doubleArray[3]);