CHANGELOG.md

0.6.1 (2023-08-27)

Fix

• relax torch version requirement

0.6.0 (2023-08-27)

Feat

• implement shifted version of frft()

0.5.4 (2023-08-18)

Fix

• ci: remove bump from pr trigger

0.5.3 (2023-08-18)

Fix

• ci: job dependency and naming

0.5.2 (2023-08-18)

Fix

• ci: combine tox and bump into build

0.5.1 (2023-08-18)

Fix

• add missing project description

0.5.0 (2023-08-16)

Feat

• layer: add non-trainable option

0.4.1 (2023-08-07)

Fix

• allow float order for frft()
• layer: add __repr__ for frft layers
• gradient error of frft() for integer orders
• gpu utilization

0.4.0 (2023-08-03)

Feat

• update sample training notebook
• layer: add discrete transform version
• add inverse wrappers for transforms
• ndim: add n-dimensional dfrft
• implement dfrft matrix
• dfrft: implement dis_s() helper function
• discrete: implement P construction
• initialize dfrft implementation

Fix

• allow Tensor type for dfrft order
• dfrft: ndim implementation
• layer: add dimension to FrFT

Refactor

• filenames due to function names
• dfrft: update naming convention
• rename fracF to frft
• change trainable-frft to torch-frft

0.3.0 (2023-07-31)

Feat

• ndim: increase ndim compatibility
• ndim: start converting fracF() to supply dim selection
• implement .mul() with nth dimension
• initialize trainable module
• convert bizdec() & bizinter() to ndim

Fix

• dimension check in einstr generation
• ndim: corefrmod2() for n-dimension
• with ndim update fracF() diff'able
• 1D tensor from (N,1) to (N,)
• remove jax based dfrt implementation

0.2.0 (2023-07-27)

Feat

• update fracF() for integer values
• torch: initialize fracF implementation
• dfrt: start importing functions for dfrt

Fix

• incorrect calculation of corefrmod2() (#2)

0.1.0 (2023-06-07)

Feat

• init: trainable frft implementation

Fix

• commitizen: project name and version