We'd like to announce the release of our code, {\sc CosmoNet} which uses neural networks to accelerate parameter estimation. We have so far trained nets to produce C_l^(TT, TE, EE) spectra and WMAP3 likelihoods over the simplest [tex]{\rm \Lambda-CDM}[/tex] cosmology -thus removing the most computationally costly element of running codes such as {\sc CosmoMC}. We will shortly release nets for a more general cosmology and program executables to enable users to train nets using their own data for any model of their choosing.
For more details see our recent submission to astro-ph/0608174
The nets and an implementation in {\sc CosmoMC} can be downloaded at:
http://www.mrao.cam.ac.uk/software/cosmonet
We would welcome any feedback from users.
Fast cosmological parameter estimation using neural networks: CosmoNet
-
- Posts: 2
- Joined: August 02 2006
- Affiliation: Cavendish Astrophysics, Cambridge