cassa 2.1.1

Python package to perform unsupervised and semi-supervised machine learning (ML) classification algorithms on generic tensors of pre-processed data

Classification with Automated Semi-Supervised Algorithms (CASSA) package

Overview

CASSA is a Python package to perform unsupervised and semi-supervised machine learning (ML) classification algorithms on generic tensors of pre-processed data, such as time series, altitude profiles, images, DDMs and spectra. Mainly tested on Earth Observation (EO) satellites data, such as GNSS-RO sTEC profiles and GNSS-R DDMs. It produces a database of labeled clusters that can be used to classify new unlabeled data. The documentation is available at https://cassa.readthedocs.io/en/latest/.

It includes the following blocks:

• Parallelized distance matrix computation using earth mover's distance (EMD, aka Wasserstein metric)
• Spetral clustering using precomputed distance matrix
• Self-tuned spectral clustering using precomputed distance matrix
• HDBSCAN clustering using precomputed distance matrix
• Classification of new data based on database of labeled clusters

Installation

pip install cassa

Authors

• Giorgio Savastano ([e-mail])
• Karl Nordstrom ([e-mail])

References

Savastano, G., K. Nordström, and M. J. Angling (2022), Semi-supervised Classification of Lower-Ionospheric Perturbations using GNSS Radio Occultation Observations from Spire’s Cubesat Constellation. Submitt. to JSWSC.