Astrophysics Group

Cavendish Laboratory

CosmoNet

Accelerated cosmological parameter estimation using Neural Networks

wlike

CosmoNet is an algorithm for accelerating cosmological parameter estimation. It uses multilayer perceptron neural networks to accurately determine CMB power spectra (temperature and polarisation), the matter transfer function and experimental likelihoods from the WMAP 3-year data. Using CAMB and WMAP codes this can take > 10 seconds per model. CosmoNet can determine these in < 50 micro seconds. We have implemented the code in a version of the popular CosmoMC sampler for convenience.

For the program, documentation and neural networks go to our downloads section.

Download the original letter astro-ph/0608174.

Download the latest paper astro-ph/0703445 which includes our latest release capable of computing both flat and non-flat cosmologies.

Web page maintained by Michael Bridges. Modified: March 2007