Skip to content

ML4GW/SageNet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

41 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SageNetGW

Overview

SageNet+ is an advanced Python package for emulating the stochastic gravitational wave background (SGWB) spectra from inflation, extending the SageNet framework described in Zhang et al. (2025). It leverages deep learning models (LSTM, Transformer, CosmicNet2, or RNN) and numerical solvers from stiffGWpy to predict the energy density spectrum with high accuracy and computational efficiency. SageNet+ supports a wide range of cosmological parameters and achieves a ~10,000-fold speedup over traditional numerical methods.

For more details, see https://github.com/YifangLuo/SageNet and https://github.com/bohuarolandli/stiffGWpy

Installation

SageNet+ is available on PyPI and can be installed using pip:

pip install sagenetgw

Dependencies

  • Python 3.8+
  • PyTorch (>=2.0.0)
  • NumPy (>=1.20.0)
  • scikit-learn (>=1.0.0)

Quick Start

Below is a simple example to predict an SGWB spectrum using SageNet+:

from sagenetgw.classes import GWPredictor
import numpy as np
from matplotlib import pyplot as plt

predictor = GWPredictor(
        model_type='Transformer',
        device="cpu"
    )

prediction = predictor.predict({
    "r":3.9585109e-05, 
    "n_t":1.0116972, 
    "kappa10":110.42477, 
    "T_re":0.17453859, 
    "DN_re":39.366618,
    "Omega_bh2":0.0223828, 
    "Omega_ch2":0.1201075, 
    "H0":67.32117, 
    "A_s":2.100549e-9
})
pred_coords = np.column_stack((prediction['f'], prediction['log10OmegaGW']))
plt.plot(pred_coords[:, 0], pred_coords[:, 1], '--', color="royalblue", marker='.')

Ensure CUDA is installed if using GPU acceleration (by device='cuda').

Parameter Ranges

The following cosmological parameters are supported:

Parameter Range Scale
r [1e-40, 1] Logarithmic
n_t [-1, 6] Linear
kappa10 [1e-7, 1e3] Logarithmic
T_re [1e-3, 1e7] GeV Logarithmic
DN_re [0, 40] Linear
Omega_bh2 [0.005, 0.1] Linear
Omega_ch2 [0.001, 0.99] Linear
H0 [20, 100] km/s/Mpc Linear
A_s [exp(1.61)/1e10, exp(3.91)/1e10] Linear

Citation

If you use SageNet+ in your research, please cite:

Zhang F, Luo Y, Li B, et al. (2025) SageNet: Fast Neural Network Emulation of the Stiff-amplified Gravitational Waves from Inflation. The Astrophysical Journal Supplement Series. 2025;279(2):44. doi:10.3847/1538-4365/ade4c6

License

SageNet+ is licensed under the MIT License. See the LICENSE file for details.

About

No description, website, or topics provided.

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages