Gaussian process regression
2016-08-23
3 0 0
4.0
Other
Earn points
The program for gaussian processs regression and classification. Wonderful working programme.
A simple one-dimensional regression exercise computed in two different ways:
- A noise-free case with a cubic correlation model
- A noisy case with a squared Euclidean correlation model
In both cases, the model parameters are estimated using the maximum likelihood principle.
The figures illustrate the interpolating property of the Gaussian Process model as well as its probabilistic nature in the form of a pointwise 95% confidence interval.
Note that the parameter nugget is applied as a Tikhonov regularization of the assumed covariance between the training points. In the special case of the squared euclidean correlation model, nugget is mathematically equivalent to a
matlab
Related Source Codes
GMSK Linear Receiver
0
0
no vote
NSGA-II algorithm
0
0
no vote
NSGA-III multi-objective optimization algorithm
0
0
no vote
Compressed sensing example
0
0
no vote
CFAR detector example
0
0
no vote
No comment