Hierarchical Gaussian Process Library
This is a simple Gaussian process library that supports different compositions of kernels.
It is designed to support inference and fit of hyperparameters for
- Composition of Squared Exponential and Linear kernels
- Mixture of kernels with linear input transformations on the data
- Regression with a linear combination of mean functions
- Scale factors can be learned along with all other kernel hyperparameters
- A simple constant prior can be learned by using a constant value (e.g. 1) as a mean function
- Zero mean constant variance noise on sampled data through the composition of a noise kernel