There are 2 parts and notebooks in this session. We will be using part1_emulator.ipynb and part2_inference.ipynb to demonstrate how one can use neural networks to emulate a complex model and how to use the emulator in Bayesian inference. The folder, answers, contains the same 2 notebooks with all blocks filled. It also has another notebook and corresponding python script for running a nested sampler, as well as the slides (no animation) used during the demonstration.
-
Required modules on Gadi:
module load tensorflow #(This also automatically load intel-mkl/2021.4.0 python3/3.10.4 cuda/11.6.1 cudnn/8.2.2-cuda11.4 nccl/2.11.4 openmpi/4.1.2) -
Required Python packages:
pip install corner emcee ultranest jupyter jupyterlab scipy mpi4py -
Download data from:
https://zenodo.org/record/8050367