Gaussian Process Regression



GPR is a non-parametric regression technique which generalizes well for sparse datasets and allows to quantify the regression error due to its Bayesian nature.

1. Click to add a data point
2. You can drag the data points
3. Drag a data point into the trash to delete it
4. Use the sliders to tune the hyperparameters
5. Click the button "Autotune" for automatic hyperparameter tuning
6. Select different kernels in the dropdown menu
7. Select number of samples to be drawn in the dropdown menu

Red line: Posterior mean (prediction)
Gray area: Posterior standard deviation (uncertainty)

by Thomas Beckers, tbeckers@seas.upenn.edu, www.tbeckers.com

Click here for an introduction to Gaussian Process Models

v1.9
Likelihood: -