View source on GitHub
|
Package for training Gaussian Processes in time less than O(n^3).
Modules
fast_gp module: Fast likelihoods etc. for Gaussian Processes.
fast_gprm module: Fast GPRM.
fast_log_det module: Functions for quickly computing approximate log det of a big PSD matrix.
fast_mtgp module: Fast likelihoods etc. for Multi-task Gaussian Processeses.
linalg module: Linear algebra routines to use for preconditioners.
linear_operator_sum module: Expresses a sum of operators.
mbcg module: Modified Batched Conjugate Gradients (mBCG) for JAX.
partial_lanczos module: Run Lanczos for m iterations to get a good preconditioner.
preconditioners module: Library of popular matrix preconditioners.
schur_complement module: The SchurComplement kernel.
Classes
class GaussianProcess: Fast, JAX-only implementation of a GP distribution class.
class GaussianProcessConfig: Configuration for distributions in the FastGP family.
class GaussianProcessRegressionModel: Fast, JAX-only implementation of a GP distribution class.
class MultiTaskGaussianProcess: Fast, JAX-only implementation of a MTGP distribution class.
class ProbeVectorType: Enum where members are also (and must be) ints
class SchurComplement: The fast SchurComplement kernel.
Functions
get_preconditioner(...): Return the preconditioner of the given type for the given matrix.