Multiclass and Multilabel Data in Vowpal Wabbit Syntax

Datasets4Vw is a collection of multi-class and multi-label datasets specifically formatted for the Vowpal Wabbit (VW) machine learning library. Created by mwydmuch and last updated in April 2018, the repository focuses on providing data optimized for extreme-large-scale classification tasks.

Use Cases

• Training multi-label classification models using Vowpal Wabbit's reduction algorithms
• Benchmarking extreme multi-label learning (XML) performance on large-scale text data
• Testing online learning throughput with VW-formatted input streams

Strengths

• Pre-formatted for immediate Vowpal Wabbit (VW) ingestion
• Supports multi-label classification targets
• Categorized for extreme-large-scale dataset benchmarking

Limitations

• Unknown license status may restrict commercial use
• Lack of descriptive metadata for individual features
• No updates or maintenance since 2018

Provenance

Source

mwydmuch (GitHub)

Freshness

The dataset is stale, with the last recorded update occurring in April 2018.