Ostrich is a library for performing emulation, also called surrogate modeling, using PCA and Gaussian process interpolation. Ostrich was used in Cromer et al, 2022 to emulate a galaxy cluster weak-lensing model, and its methodology is described there. It is based on an algorithm used in Heitmann et al, 2009 to emulate the power spectrum from cosmological simulations.
The basic premise that you have a function which is expensive to computer, so you want to create an emulator or surrogate which is trained on samples from this expensive function, and can then quickly recreate and interpolate between those samples accurately, with far less computational cost.
For now, I recommend installing using pip: run
pip install --user -e .
inside the base directory. To install dependencies, run
pip install --user -r requirements.txt
Currently there are Jupyter notebooks demoing usage in the ipython directory of the repository. In the future these will be updated and additional documentation will be created.