demu is an open-source R package implementing a Gaussian process optimal design emulator based on Determinantal point processes.
demu implements a determinantal point process emulator for probabilistically sampling optimal designs for Gaussian process (GP) regression models. Currently,
demu is a proof of concept implementation that implements basic DPP sampling, conditional DPP sampling for drawing designs of fixed size
n, sequential DPP sampling to build designs iteratively and a faster C++ implementation of the conditional DPP sampler using sparse matrices. The package supports popular stationary correlation functions commonly used in GP regression models, including the Gaussian and Wendland correlation functions.
The main model fitting functions in the package include
sim.dpp.modal() for dense correlation matrices and
sim.dpp.modal.fast() for sparse correlation matrices. These functions use a grid-based approximation to sample from the relevant DPP model.
Matthew T. Pratola <firstname.lastname@example.org> [aut, cre, cph]
Pratola, Matthew T., Lin, C. Devon, and Craigmile, Peter. (2018) Optimal Design Emulators: A Point Process Approach. arXiv:1804.02089.