Napkinxc: Extreme Multi-label Classification with Probabilistic Label Trees

Napkinxc provides implementations for extreme multi-label classification (XMLC) using Probabilistic Label Trees (PLTs) and Hierarchical Softmax (HSM). Developed by mwydmuch and updated in November 2025, the project focuses on high-performance categorization for large label spaces.

Use Cases

• Training models for extreme multi-label classification using the Probabilistic Label Trees (PLT) feature
• Reducing computational complexity in large-scale categorization via the Hierarchical Softmax (HSM) implementation
• Developing multi-class classification pipelines using the provided Python interface

Strengths

• MIT license for flexible redistribution
• Implements specialized Probabilistic Label Trees (PLTs)
• Maintained with updates as of November 2025

Limitations

• Input provides no specific data records or row counts as it is a software repository
• Lack of sample data makes immediate verification of model performance difficult without external sources

Provenance

Source

mwydmuch

Freshness

Last updated November 2025.