CoLFI: Cosmological Likelihood-Free Inference
Abstract
CoLFI (Cosmological Likelihood-Free Inference) estimates parameters directly from the observational data sets using neural density estimators (NDEs); it is a fully ANN-based framework that differs from the Bayesian inference. The package contains three NDEs that are used to estimate parameters: an artificial neural network (ANN), a mixture density network (MDN), and a mixture neural network (MNN). CoLFI can learn the conditional probability density using samples generated by models, and the posterior distribution can be obtained for given observational data.
- Publication:
-
Astrophysics Source Code Library
- Pub Date:
- September 2023
- Bibcode:
- 2023ascl.soft09006W
- Keywords:
-
- Software