normalizing-flows

This repo contains the code used to run a forward modeling lensing analysis on a population of subhalos generated from a normalizing flows algorithm. The code is split into 3 main steps:

1. Generate a set of Galacticus merger trees. The file darkMatterOnlySubHalosPipeline.xml is the input parameter file that gets read into Galacticus, dmosh_pipeline.sh is the job script used to run Galacticus on this parameter file, and darkMatterOnlySubHalosPipeline.hdf5 is the output file after Galacticus has generated the merger trees.
2. Train the emulator on Galacticus data. The script normalizing_flows_pipeline.py reads in the Galacticus data from step 1, and feeds the data into the emulator. flows_pipeline.sh is the job file used to run normalizing_flows_pipeline.py.
3. Perform the forward modeling analysis to generate $S_\text{lens}$ values. The script subhalos_inference_pipeline.py uses the emulator to generate a population of emulated subhalos, and runs it through a forward modeling analysis from the analysis code samana. The .sh file used to run this final step is inference_pipeline.sh. Currently, the subhalos_inference_pipeline.py file is set up to produce $S_\text{lens}$ values from an empirical model. To produce a population of emulated subhalos, in line 140 set the variable preset_model_name = DMEmulator instead of preset_model_name = WDM. Also, uncomment line 152, as this is an extra input that gets read into the kwargs_sample_realization dictionary when running the emulator.

NOTE1: In steps 2 and 3, both files normalizing_flows_pipeline.py and subhalos_inference_pipeline.py import functions/classes from the script emulator_pipeline.py. This script contains functions for transforming Galacticus to and from it's normalized coordinates and hypercube coordinates, as well as the code which builds the normalizing flows architecture.

NOTE2: When performing step 3, the code requires certain files of pyHalo which have been modified to incorporate the emulator in the forward modeling analysis. The first file is pyHalo/preset_models.py, and the second file is pyHalo/PresetModels/external.py.