17.06.2024

New article published in the "Journal of Open Source Software"

Title: "cblearn: Comparison-based Machine Learning in Python" by David-Elias Künstle and Ulrike von Luxburg

Abstract: 
The cblearn package implements comparison-based machine learning algorithms and routines for processing comparison-based data in Python.
Comparison-based learning algorithms are used when only comparisons of similarity between data points are available, but no explicit similarity scores or features. For example, humans have difficulty assigning numerical similarities to apples, pears, and bananas. However, they can easily compare the similarity of pears and apples to the similarity of apples and bananas-pears and apples usually appear more similar.
Comparison-based algorithms exist for most machine learning tasks, such as clustering, regression, classification, and the most popular, ordinal embedding.

cblearn provides an ecosystem for comparison-based data, with access to several real-world datasets and a collection of algorithm implementations. The package is fast and easy to use for applications, but flexible for developing new algorithms and methods. cblearn integrates well with the scientific Python ecosystem and has been used for algorithm development and data analysis in several studies.

To see the whole article, please visit our publications page.