STAR_emu

Stellar-to-halo mass Relation Emulator for CAMELS-IllustrisTNG simulations.

A fast polynomial ridge-regression emulator that predicts the stellar–halo mass relation (SHMR) as a function of 6 cosmological and astrophysical parameters, trained on 1,000 Latin Hypercube simulations from the CAMELS project.

Quick start

from star_emu import SHMREmulator
import numpy as np

emu = SHMREmulator()

log_mhalo = np.linspace(10.5, 13.5, 50)
log_mstar = emu.predict(log_mhalo,
                        Omega_m=0.3, sigma_8=0.8,
                        A_SN1=1.0, A_AGN1=1.0,
                        A_SN2=1.0, A_AGN2=1.0)

Parameters

Parameter	Description	Range
Ω_m	Matter density	0.1 – 0.5
σ_8	Clustering amplitude	0.6 – 1.0
A_SN1	Supernova feedback strength	0.25 – 4.0
A_AGN1	AGN feedback strength	0.25 – 4.0
A_SN2	Supernova feedback speed	0.5 – 2.0
A_AGN2	AGN feedback speed	0.5 – 2.0

Files

File	Description
star_emu.py	Emulator module (SHMREmulator class)
extract_shmr.py	Data extraction script (requires access to CAMELS HDF5 catalogs)
data/emulator_model.npz	Pre-trained model weights
data/shmr_data.npz	Extracted SHMR data from 1,000 simulations
emulator_lab.ipynb	Interactive student lab notebook (~2 hr)
emulator_lab_solutions.ipynb	Solutions notebook
emulator_tutorial.ipynb	Reference tutorial notebook

Requirements

• Python ≥ 3.8
• numpy
• matplotlib (for notebooks)

Lab notebook

The emulator_lab.ipynb notebook is a self-contained ~2 hour lab session designed for students with no prior machine-learning experience. It builds up from scatter plots and best-fit lines to polynomial regression and emulator construction, with guided exercises and challenge problems.