Code and plots for the NST Physics Part III project "Astronomical Linear Simulation-based Inference"
Web reference: Galileo - Part III Projects supervised by Will Handley
We develop the theoretical framework of Linear Simulation-based Inference (LSBI), an application of likelihood-free inference where the model is approximated by a linear function of its parameters and the noise is assumed to be Gaussian with zero mean. We obtain analytical expressions for the posterior distributions of hyperparameters of the linear likelihood in terms of samples drawn from a simulator. This method is applied to several toy models, and to emulated datasets for the Cosmic Microwave Background power spectrum and the sky-averaged 21cm hydrogen line. We find that convergence is achieved after
- Python 3.9+
- lsbi - GitHub, Documenation
- globalemu - GitHub, Documenation
- cmbemu - GitHub
- getdist - GitHub, Documentation
Handley et al, (2024) lsbi: Linear Simulation Based Inference.
Bevins, H., Handley, W. J., Fialkov, A., Acedo, E. D. L., and Javid, K. (2021). GLOBALEMU: A novel and robust approach for emulating the sky-averaged 21-cm signal from the cosmic dawn and epoch of reionisation. arXiv:2104.04336
Lewis, A. (2019). GetDist: a Python package for analysing Monte Carlo samples. arXiv preprint arXiv:1910.13970.