Metadata-Version: 2.3
Name: waveletnn
Version: 0.2.1
Summary: WaveletNN blocks and losses
License: MIT
Keywords: wavelet,dwt,idwt,pytorch
Author: Scurrra (Ilja Baroŭski)
Author-email: iscurrra@gmail.com
Requires-Python: >=3.11
Classifier: Programming Language :: Python :: 3
Classifier: Development Status :: 4 - Beta
Requires-Dist: numpy
Requires-Dist: torch (>=2)
Project-URL: Issues, https://github.com/Scurrra/WaveletNN-PyTorch/issues
Project-URL: Repository, https://github.com/Scurrra/WaveletNN-PyTorch.git
Description-Content-Type: text/markdown

# Wavelet Neural Networks

[![PyPI - Version](https://img.shields.io/pypi/v/waveletnn?style=flat)](https://pypi.org/project/waveletnn/)
[![GitHub License](https://img.shields.io/github/license/Scurrra/WaveletNN-PyTorch?style=flat)](https://github.com/Scurrra/WaveletNN-PyTorch/tree/master?tab=MIT-1-ov-file)


Implementation of orthonormal and biorthogonal wavelet transforms via convolutions. Multibatch single-channel one- and -two-dimensional data is supported. For analysis kernels of even length are supported, while for inverse transform kernels are required to have length `4k + 2`. 

Package provides loss functions for wavelet's kernels regularizations to preserve features of both orthonormal and biorthogonal wavelets. 
