Example on 2-d data

import openturns as ot
import otsliced
import matplotlib.pyplot as plt

Create 2-d data X and 1-d feature Y

N = 100
X = ot.Normal([0.0] * 2, [0.1] * 2).getSample(N)
X += [[-i / (N - 1), 2 * i / (N - 1)] for i in range(N)]
X = X - X.computeMean()
f = ot.SymbolicFunction(["x1", "x2"], ["4*(x1+2*x2)+2"])
Y = f(X) + ot.Normal(0.0, 0.2).getSample(N)

Plot data

plt.scatter(X.getMarginal(0), X.getMarginal(1), c=Y)
plt.xlabel('x1')
plt.ylabel('x2')
plt.axis('square')
plt.title("2D colored Dataset")

Text(0.5, 1.0, '2D colored Dataset')

Run the SIR algorithm

algo = otsliced.SlicedInverseRegression(X, Y)
algo.run()
transformation = algo.getResult().getTransformation()

Show SIR direction

plt.scatter(X.getMarginal(0), X.getMarginal(1), c=Y)
plt.xlabel('x1')
plt.ylabel('x2')
plt.axis('square')
plt.title("First direction of SIR")
dir_sir = algo.getResult().getDirections()
plt.plot([-dir_sir[0, 0] * 30, dir_sir[0, 0] * 30], [-dir_sir[1, 0] * 30, dir_sir[1, 0] * 30], c='red')

[<matplotlib.lines.Line2D object at 0x7f6036065b10>]

plt.show()