5
$\begingroup$

I am trying to use Gaussian Processes for fitting smooth functions to some datapoints. I am using scikit-learn library for python and in my case my input are two dimensional spatial coordinates and the output are some transformed version and also 2-D spatial coordinates. I generated some dummy test data and tried to fit a GP model to it. The code that I used was as follows:

from sklearn.gaussian_process import GaussianProcessRegressor
from sklearn.gaussian_process.kernels import RBF, ConstantKernel as C
import numpy as np

# Some dummy data
X = np.random.rand(10, 2)
Y = np.sin(X)

# Use the squared exponential kernel
kernel = C(1.0, (1e-3, 1e3)) * RBF(10, (1e-2, 1e2))
gp = GaussianProcessRegressor(kernel=kernel, n_restarts_optimizer=9)
# Fit to data using Maximum Likelihood Estimation of the parameters
gp.fit(X, Y)
print(X)
# Evaluate on a test point
test = np.random.rand(1, 2)
test[:, 0] = 1.56
test[:, 1] = 0.92
y_pred, sigma = gp.predict(test, return_std=True)
print(test, np.sin(test))  # The true value
print(y_pred, sigma)  # The predicted value and the STD

I was wondering if there is a good way to visualize the model fit. As my input and output dimensions are both 2-D, I am not sure how I can visualize it quickly so that I get an idea of the model fit (particularly want to know the smoothness and variance of the model prediction between the points). Most examples online are, of course, for 1-D case.

$\endgroup$
5
$\begingroup$

I think a good approach in your case could be to

  1. Fit the multivariate GP model on a few training points, as you do now
  2. Take advantage of the fact you have the ground truth function in order to generate true values and predicted values for a range of inputs.
  3. Plot comparisons of the "marginal" and "joint" outputs for these ranges of values.

Preparing 2-D inputs as a Matlab-style meshgrid:

delta = 0.025
x = np.arange(-1, +1, delta)
y = np.arange(-1, +1, delta)
X, Y = np.meshgrid(x, y)

Generating predictions from the fitted GP model for all the combinations of 2-D X inputs, and then separating the 2-D outputs into individual arrays for later use:

test = np.stack([np.ravel(X), np.ravel(Y)], axis=1)
y_pred, sigma = gp.predict(test, return_std=True)
y_pred_fromX = y_pred[:,0].reshape(X.shape)
y_pred_fromY = y_pred[:,1].reshape(X.shape)

For the first dimension of the 2-D output, we plot the actual & predicted values as contours, with the axes representing the 2-D inputs:

import matplotlib.pyplot as plt

plt.figure(figsize=(10,5))
plt.subplot(121)
plt.contour(X, Y, np.sin(X), 20)
plt.title('1st dim: True')
plt.subplot(122)
plt.contour(X, Y, y_pred_fromX, 20)
plt.title('1st dim: Predicted')

enter image description here

Same for the second dimension of the 2-D output:

plt.figure(figsize=(10,5))
plt.subplot(121)
plt.contour(X, Y, np.sin(Y), 20)
plt.title('2nd dim: True')
plt.subplot(122)
plt.contour(X, Y, y_pred_fromY, 20)
plt.title('2nd dim: Predicted')

enter image description here

Focussing on the 2-D output alone, scatterplots of joint occurences are not particularly helpful. Here the axes are the 2-D output values:

plt.figure(figsize=(10,5))
plt.subplot(121)
plt.scatter(np.sin(X), np.sin(Y))
plt.title('True: scatterplot')
plt.subplot(122)
plt.scatter(y_pred_fromX, y_pred_fromY)
plt.title('Predicted: scatterplot')

enter image description here

But Seaborn's jointplots are much more useful. Once again, axes are 2-D output values, and the plot represents a calculated density:

import seaborn as sns

plt.figure()
sns.jointplot(x=np.sin(X), y=np.sin(Y), kind='kde')
plt.title('True: jointplot')
plt.figure()
sns.jointplot(x=y_pred_fromX, y=y_pred_fromY, kind='kde')
plt.title('Predicted: jointplot')

enter image description here enter image description here

$\endgroup$
1
  • $\begingroup$ Great suggestion. I ended up doing the plots for each dimension separately as well. The jointplot idea is great. I never used seaborn before but it looks like it is worthy of a look! $\endgroup$ – Luca Feb 6 '18 at 15:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.