Paper & citation

Paper & citation#

pdft is the reference implementation accompanying

Shiwen An, Zhongyi Ni, Huanhai Zhou, Jin-Guo Liu. Fast Trainable Multilinear Bases for Image Compression. arXiv:2608.00053 (2026). arxiv.org/abs/2608.00053

Citing pdft#

If you use this package in your research, please cite:

@misc{an2026fast,
  title         = {Fast Trainable Multilinear Bases for Image Compression},
  author        = {An, Shiwen and Ni, Zhongyi and Zhou, Huanhai and Liu, Jin-Guo},
  year          = {2026},
  eprint        = {2608.00053},
  archivePrefix = {arXiv},
  primaryClass  = {eess.IV},
  url           = {https://arxiv.org/abs/2608.00053},
}

Background reading#

Notes from the upstream Julia package, at the pinned commit:

Relation to ParametricDFT.jl#

pdft ports ParametricDFT.jl and is pinned to upstream commit a201a27 (pdft.__upstream_ref__). Matching Julia’s behavior is the package’s main correctness criterion: the test suite compares Python output against goldens generated by the Julia reference. The Python port is MIT-licensed, as is ParametricDFT.jl (© 2025 nzy1997).