- GP regression with RBF and polynomial kernels
- A graphical illustration of the online learning
- Projection in the feature space
- The evolution of the KL-divergences for the online training
- Parameter decomposition
- Errors made by ignoring the covariance-term in the score
- A graphical illustration of the expectation propagation algorithm
- Regression results for the
function.
- Regression results for the Friedman dataset.
- Regression results for the Boston dataset.
- Sparse EP applied to Crab data.
- Sparse EP applied to Sonar data.
- Results for the USPS dataset for binary and combined
classification.
- Multiclass classification with rejecting the uncertain predictions.
- The kernel used for density estimation.
- Results for the GP density estimation.
- The local ``likelihoods'' in the wind field model.
- Updates for the vectorial GP
- NWP wind-field prediction and sparse GP predictions.
- Computing the relative weight of a GP approximation
- The predicted wind-fields based on the sparse GP.
- Grouping of the GP parameters (Fig 3.3 repeated).
L CSATO
2003-04-23