RandomFeatures
A julia package to construct and apply random feature methods for regression. RandomFeatures can be viewed as an approximation of kernel methods. They can be used both as a substitution in Kernel ridge regression and Gaussian Process regresison.
| Module | Purpose |
|---|---|
| RandomFeatures | Container of all tools |
| Samplers | Samplers for constrained probability distributions |
| Features | Builds feature functions from input data |
| Methods | Fits features to output data, and prediction on new inputs |
| Utilities | Utilities to aid batching, and matrix decompositions |
Highlights
- A flexible probability distribution backend with which to sample features, with a comprehensive API
- A library of modular scalar functions to choose from
- Methods for solving ridge regression or Gaussian Process regression problem, with functions for producing predictive means and (co)variances using fitted features.
- Examples that demonstrate using the package
EnsembleKalmanProcesses.jlto optimize hyperparameters of the probability distribution.
Authors
RandomFeatures.jl is being developed by the Climate Modeling Alliance. The main developers are Oliver R. A. Dunbar and Thomas Jackson, with acknowledgement that the code was based on a python repository developed by Oliver R. A. Dunbar, Maya Mutic, and Nicholas H. Nelsen.