FOSCAT
FOSCAT (Field cOmpression via Scattering CovariAnce Transform) is a Python library for wavelet/scattering-based statistical analysis, synthesis, denoising, and deep learning on 2D fields and on the sphere with HEALPix pixelisation.
The central idea: any spatial field can be summarised by a compact vector of scattering-covariance statistics Φ(d) — encoding multi-scale, cross-orientation correlations beyond what a power spectrum captures. FOSCAT can then synthesise a new field matching those statistics, use them as a loss for denoising or component separation, or feed them into neural-network architectures that operate directly on spherical geometry.
Install
pip install foscat
Full environment for running the example notebooks:
micromamba create -n foscat python=3.10
micromamba activate foscat
pip install foscat torch healpy xarray gcsfs zarr jupyterlab matplotlib scipy
Backend note: FOSCAT uses PyTorch as its only active backend (TensorFlow and NumPy backends are no longer maintained).
Minimal HEALPix synthesis
import numpy as np
import foscat.scat_cov as sc
from foscat.Synthesis import Loss, Synthesis
nside = 64
target_map = np.random.randn(12 * nside**2)
# build the scattering operator and compute target statistics
scat_op = sc.funct(KERNELSZ=5, NORIENT=4, nstep_max=4, all_type='float64')
target_stat = scat_op.eval(target_map)
# define a differentiable loss
def synth_loss(x, scat_op, args):
ref = args[0]
stat = scat_op.eval(x)
return stat.reduce_mean_batch((stat - ref) ** 2)
loss = Loss(synth_loss, scat_op, target_stat)
solver = Synthesis([loss], eta=0.03)
x0 = np.random.randn(12 * nside**2)
result = solver.run(x0, NUM_EPOCHS=300, EVAL_FREQUENCY=10)
What FOSCAT can do
| Capability | Entry point | Example notebook |
|---|---|---|
| HEALPix synthesis | foscat.scat_cov.funct + Synthesis |
Demo_Synthesis.ipynb |
| 2D image synthesis | foscat.scat_cov2D.funct + Synthesis |
Synthesis2D.ipynb |
| Denoising | two-loss synthesis (statistics + fidelity) | Denoising-2D.ipynb |
| Component separation | multi-loss synthesis | Remove_CMB.ipynb |
| Local spherical analysis | cell_ids in FoCUS / HealpixUNet |
local_foscat.ipynb |
| Spherical U-Net (regression/segmentation) | HealpixUNet |
CNN_local.ipynb, CNN_ecmwf.ipynb |
Example notebooks are maintained in demo-foscat-pangeo-eosc.
Documentation
Full documentation (installation, mathematical background, API reference, user guide) is published at:
https://jmdelouis.github.io/FOSCAT/
To build the docs locally:
cd docs_sphinx
make install # pip install requirements.txt
make html # → _build/html/index.html
make livehtml # live-reload on save (requires sphinx-autobuild)
Citation
If you use FOSCAT in a publication, please cite:
@software{foscat,
author = {Delouis, Jean-Marc and Foulquier, Theo},
title = {{FOSCAT}: Field cOmpression via Scattering CovariAnce Transform},
year = {2026},
url = {https://github.com/jmdelouis/FOSCAT},
version = {2026.04.1},
}
The underlying statistical framework is described in:
- Allys et al. (2019), Rwst, a comprehensive statistical description of the non-Gaussian structures in the ISM, A&A 629, A115.
- Allys et al. (2020), New interpretable statistics for large-scale structure analysis and generation, Phys. Rev. D 102, 103506.
License
BSD 3-Clause. See LICENSE.
Metadata
Release files for foscat 2026.6.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| foscat-2026.6.1.tar.gz | 224.5 kB | Without provenance | Details for foscat-2026.6.1.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| foscat-2026.6.1-py3-none-any.whl | Python 3 | none | any | Details for foscat-2026.6.1-py3-none-any.whl |
Total release size: 468.7 kB
Release files / foscat-2026.6.1.tar.gz
| Download URL | foscat-2026.6.1.tar.gz |
|---|---|
| Size | 224.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
3d2518fca343143085a51acfd5176433ca07713ea0e9d5b80307cd696737131e
|
|
BLAKE2b-256 checksum How to use checksums |
336626a30f0c1327dc9419013a34c1af411040b1126b725556d78e14ea2f9a43
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.11.13
|
Release files / foscat-2026.6.1-py3-none-any.whl
| Download URL | foscat-2026.6.1-py3-none-any.whl |
|---|---|
| Size | 244.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
f3f32f4b7f4a481345f493e57d38acbc0a827df9fc510af95a7d03d5ea6093c3
|
|
BLAKE2b-256 checksum How to use checksums |
65cbaf68f5797e8bed885b5436bc1aa25236700930786ff5881298d7dcc1dbbc
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.11.13
|
Security for foscat 2026.6.1
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.