https://github.com/tmcclintock/boltzmann_emulator

A first stab at emulating the output of Boltzmann codes for power spectra.
https://github.com/tmcclintock/boltzmann_emulator

Last synced: 24 days ago
JSON representation

A first stab at emulating the output of Boltzmann codes for power spectra.

• Host: GitHub
• URL: https://github.com/tmcclintock/boltzmann_emulator
• Owner: tmcclintock
• License: mit
• Created: 2019-01-08T18:19:13.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2019-02-11T13:46:22.000Z (almost 6 years ago)
• Last Synced: 2024-12-14T17:09:21.094Z (about 2 months ago)
• Language: OpenEdge ABL
• Homepage:
• Size: 2.25 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Boltzmann emulator

Boltzmann codes such as CAMB or CLASS can be very slow. This repository contains a tool for emulating the outputs of a given boltzmann code in order to estimate their outputs quickly.

## Installation

One can install the requirements with `pip install -r requirements.txt`. If you have trouble with this, it is almost certainly with the package [george](https://george.readthedocs.io/en/latest/), used for the Gaussian processes. Feel free to raise an issue if this is the case.

With the requirements installed, install this package with `python setup.py install`. Run the unit tests just by typing `pytest` in the main directory of this project. If any tests fail, please raise an issue.

## Usage

A minimal example of this package would consist of the following:

```python

import boltzmann_emulator as BE

#A 2D array of cosmological parameters

parameters = ...

#A list of redshifts

redshifts = ...

#A list of wavenumbers

k = ...

#A 2D array (N_cosmologies x (N_z x N_k)) of the power spectra

#Note that the power spectra have been stacked in the last dimension

Pkz_training = ...

#Create the emulator

Emu = BE.boltzmulator(parameters, redshifts, k, Pkz_training)

#Predict at a new location in parameter space

Pkz_predicted = Emu.predict(test_parameters)

```