public void Mean()
{
var series1 = new NDArray_Legacy <double>();
series1.Data = new double[] { 1, 2 };
var series2 = new NDArray_Legacy <double>();
series2.Data = new double[] { 3, 4 };
var np = new NDArray_Legacy <NDArray_Legacy <double> >();
np.Data.Add(series1);
np.Data.Add(series2);
//Assert.IsTrue(Enumerable.SequenceEqual(np.Mean().Data, new double[] { 2.5 }));
Assert.IsTrue(Enumerable.SequenceEqual(np.Mean(0).Data, new double[] { 2, 3 }));
Assert.IsTrue(Enumerable.SequenceEqual(np.Mean(1).Data, new double[] { 1.5, 3.5 }));
}
public static NDArray_Legacy <double> Std(this NDArray_Legacy <NDArray_Legacy <double> > np, int axis = -1)
{
var std = new NDArray_Legacy <double>();
var mean = np.Mean(axis);
// axis == -1: DEFAULT; to compute the standard deviation of the flattened array.
if (axis == -1)
{
var sum = np.Data.Select(d => d.Data.Select(p => Math.Pow(Math.Abs(p - mean.Data[0]), 2)).Sum()).Sum();
std.Data.Add(Math.Sqrt(sum / np.Size));
}
// to compute mean by compressing row and row
else if (axis == 0)
{
double[] sumVec = new double[np.Data[0].Length];
for (int d = 0; d < np.Length; d++)
{
for (int p = 0; p < np.Data[0].Length; p++)
{
sumVec[p] += np.Data[d][p];
}
}
for (int d = 0; d < np.Data[0].Length; d++)
{
mean.Data.Add(sumVec[d] / np.Length);
}
}
else if (axis == 1)
{
for (int d = 0; d < np.Length; d++)
{
double rowSum = 0;
for (int p = 0; p < np.Data[0].Length; p++)
{
rowSum += np.Data[d][p];
}
mean.Data.Add(rowSum / np.Data[0].Length);
}
}
return(std);
}