What’s New

Version 0.3.0

With this version we start a change log. We will follow the way scikit-learn does it by having a whats_new folder in the docs and a changelog file for each minor (X.Y) version. This document will be required to be updated with every pull request. For this release we will mostly use the PR descriptions, but in the future better descriptions will be required.

New contributions should be prepended to the appropriate version’s file.


  • Maintenance Fixed the issues that made Jenkins CI runs fail. #810 by Joakim Anden

  • Documentation Remove warning from sklearn example.  #828

  • Fix Removed redundant default value for alpha. #814

  • Maintenance remove size_scattering kwarg from scattering1d #837

  • Documentation updated contributing guidelines #813

  • Maintenance decouple scattering_filter_factory from ScatteringBase1D.build #842

  • Maintenance simplify compute_params_filterbank, remove _move_one_dyadic_step #848

  • Maintenance remove calibrate_scattering_filters #850

  • Maintenance Simplify filter_bank.py #833

  • Maintenance avoid passing log2_T, pad, unpad to 1D core #860

  • Maintenance Remove vectorize + write frontend-agnostic runtime checks #861

  • Maintenance Remove out_type from 1D core, perform formatting in the frontends #862

  • Documentation fixing example, now import TorchBackend3D to utilize compute_integrals #829

  • Maintenance removed useless apply function #830

  • Documentation Use tight_layout to prevent plot axes from overlapping one another #827

  • Documentation fix reconstruct_torch.py example by calling detach #826

  • Maintenance Make Scattering1D compatible with TF>=2.5 #882

  • API Change precompute_size_scattering now returns a tuple #878

  • Maintenance HarmonicScattering3D now outputs real coefficients in Numpy.  #889

  • Maintenance make NumPy the default frontend #890

  • Maintenance, STY Various ASV maintenance, style changes.  #845

  • Feature Make Q2 customizable #854

  • Maintenance deprecate self.N and self.J_pad in 1D #863

  • Maintenance make Scattering2D compatible with TF>=2.5 #883

  • Maintenance move core 1D asserts to _check_filterbanks in ScatteringBase #876

  • Maintenance check oversampling in 1D #887

  • Maintenance Deprecate self.N in 2D #886

  • Fix examples/plot_filters.py #815

  • Testing add summary call to keras 1d and 2d tests.  #888

  • Documentation Enable .to(device)` semantics for pytorch examples #874

  • Maintenance redefine cdgmm3d to call cdgmm in torch skcuda backend #865

  • Documentation changes to documentation of output shapes #820

  • Testing Add seed to 2D FFT test in TF #896

  • Maintenance update GitHub workflows #892

  • Enhancement Jax backend #757

  • Fix Update examples/1d/plot_filters.py #899

  • Fix include ‘n’ field in 1D/2D list + remove singleton dimension in 1D dict #900

  • Maintenance deprecate average in Scattering1D #897

  • Documentation document N and J in scattering_filter_factory #901