XASDenoise

XASDenoise is a Python package for denoising and preprocessing X-ray Absorption Spectroscopy (XAS) data.
It implements a data preprocessing method called stationarity warping, which aims at mitigating signal non-stationarity. The package also includes advanced denoising techniques based on Gaussian Processes and convolutional autoencoders. The package comes with ready-to-run Jupyter notebook examples and example XAS datasets.

---

Features

• Gaussian Process-based denoising for probabilistic signal recovery
• Convolutional autoencoder training and denoising pipeline utilizing autoencoder architecture
• Stationarity warping to mitigate signal non-stationarity and improve performance of any denoising method
• Preprocessing utilities for XAS data
• Example datasets and Jupyter notebooks included

---

Installation

pip install git+https://github.com/tomasaidukas/XASDenoise.git

or

git clone https://github.com/tomasaidukas/XASDenoise.git
cd XASDenoise
pip install -e .

Examples & Data

The repository includes:

• examples/ — Jupyter notebooks:

  • example0_spectrum_object_and_preprocessing.ipynb — creating and processing spectrum objects
  • example1_regular_denoiser.ipynb — basic denoising
  • example2_gaussian_process_denoiser.ipynb — Gaussian Process denoising
  • example3_autoencoder_denoiser.ipynb — autoencoder denoising
  • example4_time_resolved_data_denoising.ipynb — time-resolved spectra denoising

• examples/data/ — example XAS datasets

---

Requirements

• Python >= 3.11
• See requirements.txt for core dependencies

---

License

This project is licensed under the MIT License — see the LICENSE file for details.