Module Map#
Complete index of FOSCAT modules, classes, and key functions.
Scattering operators#
Module |
Entry point |
Description |
|---|---|---|
|
|
HEALPix / full-sphere scattering-covariance operator. Returns a |
|
|
2D planar scattering-covariance operator. |
|
|
1D scattering-covariance operator. |
|
|
Core class: wavelet filter construction, GPU dispatch, stencil management. |
Synthesis#
Module |
Entry point |
Description |
|---|---|---|
|
|
L-BFGS-B optimisation loop for synthesis / reconstruction. |
|
|
Wraps a user loss function and a scattering operator. |
Statistics objects#
Module |
Class |
Description |
|---|---|---|
|
|
Container for S0–S4 coefficient arrays with arithmetic and serialisation. |
|
|
2D variant of the statistics object. |
Neural networks#
Module |
Class |
Description |
|---|---|---|
|
|
PyTorch U-Net on HEALPix geometry with oriented spherical convolutions. |
|
|
Graph-convolutional network using FOSCAT scattering layers. |
|
|
Flat CNN on HEALPix (simple baseline). |
|
|
Legacy U-Net (TensorFlow-era, kept for compatibility). |
|
|
Vision Transformer on HEALPix tokens. |
|
|
ViT with skip connections. |
|
|
ViT on equirectangular grids. |
|
— |
2D U-Net parameterised from a HEALPix model. |
Spherical geometry#
Module |
Class / function |
Description |
|---|---|---|
|
|
Oriented spherical convolution on HEALPix neighbour stencils. |
|
|
High-level oriented convolution wrapper. |
|
|
Geodesic downsampling between HEALPix resolutions. |
|
|
Geodesic upsampling between HEALPix resolutions. |
|
— |
Bilinear interpolation between HEALPix resolutions. |
|
— |
Spline interpolation on HEALPix. |
|
— |
Circular spline interpolation. |
|
— |
1D spline interpolation utilities. |
Spherical harmonic analysis#
Module |
Key functions |
Description |
|---|---|---|
|
|
Full-sky spherical harmonic decomposition. |
|
— |
Alm analysis on lat/lon grids. |
|
— |
Local spherical harmonic analysis on regional domains. |
|
— |
Optimised local alm routines. |
Backends#
Module |
Class |
Description |
|---|---|---|
|
|
Abstract base class for all backends. |
|
|
Active PyTorch backend (GPU + CPU). |
|
|
Deprecated. TensorFlow backend (not maintained). |
|
|
Deprecated. NumPy backend — read-only; cannot run |
Loss backends (internal)#
Module |
Description |
|---|---|
|
Gradient computation via PyTorch autograd. |
|
Deprecated. Gradient computation via TF GradientTape. |
Visualisation#
Module |
Key functions |
Description |
|---|---|---|
|
|
Convenience plotting functions (HEALPix Mollweide, statistics curves). |
xarray integration#
Module |
Description |
|---|---|
|
Registers a |
|
Statistics computations integrated with xarray labels. |
|
Parameter management for xarray workflows. |