public void TrainDistributedMachines()
{
// set parameters
for (int i = 0; i < _agents.Count; i++)
{
_agents[i].GetMachine()._lamda = 0.8;
}
for (int i = 0; i < _agentNum; i++)
{
for (int j = 0; j < _agentNum; j++)
{
if (_adjacency[i, j] == 1)
{
_agents[i].AddNeighbor(_agents[j]);
}
}
}
for (int t = 0; t < _iterationTime; t++)
{
for (int i = 0; i < _agentNum; i++)
{
_agents[i].SetValue(_dataGnr.Calc(_agents[i].GetPos()));
}
for (int i = 0; i < _agentNum; i++)
{
_agents[i].Train();
}
}
}
public OneDimDataTester(KernelRegression machine = null)
{
_dataGnr = new OneDimDataGenerator();
_rnd = new Random();
_trainDataX = new double[_trainDataSize][];
_trainDataY = new double[_trainDataSize];
for (int i = 0; i < _trainDataSize; i++)
{
_trainDataX[i] = new double[1];
_trainDataX[i][0] = 2 * _rnd.NextDouble() - 1.0;
}
_trainDataY = _dataGnr.Calc(_trainDataX);
if (machine == null)
{
_regressionMachine = new KernelRegression();
}
else
{
_regressionMachine = machine;
}
}
public DistributedOneDimDataTrainer()
{
_agentNum = _trainDataSize;
_adjacency = new int[_agentNum, _agentNum];
_dataGnr = new OneDimDataGenerator();
_rnd = new Random();
_trainDataX = new double[_trainDataSize][];
_trainDataY = new double[_trainDataSize];
_agents = new List <KernelAgent>();
_centerMachine = new KernelRegression();
for (int i = 0; i < _agentNum; i++)
{
_trainDataX[i] = new double[1];
_trainDataX[i][0] = 2 * _rnd.NextDouble() - 1.0;
_agents.Add(new KernelAgent(_trainDataX[i]));
}
_trainDataY = _dataGnr.Calc(_trainDataX);
Init();
}
public void Test()
{
double[][] testDataX = new double[_testDataSize][];
double[] testDataY = new double[_testDataSize];
double[] testDataRefY = new double[_testDataSize];
for (int i = 0; i < _testDataSize; i++)
{
testDataX[i] = new double[1];
testDataX[i][0] = -1.0 + (double)((2 * i)) / (double)(_testDataSize);
}
testDataY = _dataGnr.Calc(testDataX);
testDataRefY = _regressionMachine.Predict(testDataX);
DataAnalyzer analyzer = new DataAnalyzer();
double[] err = analyzer.GetDifference(testDataY, testDataRefY);
double mean = analyzer.Mean(err);
double variance = analyzer.Variance(err);
Console.WriteLine("testDataX:");
for (int i = 0; i < testDataX.Length; i++)
{
Console.Write(testDataX[i][0] + ",");
}
Console.Write("\n");
Console.Write("\n");
Console.WriteLine("testDataY:");
for (int i = 0; i < testDataY.Length; i++)
{
Console.Write(testDataY[i] + ",");
}
Console.Write("\n");
Console.Write("\n");
Console.WriteLine("testDataRefY:");
for (int i = 0; i < testDataRefY.Length; i++)
{
Console.Write(testDataRefY[i] + ",");
}
Console.Write("\n");
Console.Write("\n");
Console.WriteLine("testDataError:");
for (int i = 0; i < err.Length; i++)
{
Console.Write(err[i] + ",");
}
Console.Write("\n");
Console.Write("\n");
Console.WriteLine("Mean: " + mean);
Console.WriteLine("Variance: " + variance);
}