Clone this wiki locally
Here are some examples of using vowpal-wabbit
- RCV1 example. Here we use online gradient descent on TFIDF transformed features while taking advantage of a cache file.
- Malicious URL example. A sequential binary classification problem with temporally correlated data.
- VW has a test suite.
less +/^__DATA test/RunTestsprovides many little examples with input data-sets and expected reference outputs.
- Matrix factorization example
- Multi Class Examples
- Weighted All Pairs (wap) multi class example
- Truncated gradient descent example
- Contextual Bandit Example
- daemon example - how to run vw in daemon mode and get predictions
- Using active_interactor.py an interactive way to interface with vw's active learning
- Using vw-hypersearch a vw wrapper to search for "optimal" (argmin) parameters
- using vw varinfo - a vw wrapper to inspect your training-set & model
- Using vw regr - a script for aiding regression applications
- vw-top-errors: online learning debugging for better models
The source tree has a sub-directory
demo which includes somewhat more involved examples:
- DNA splice Site recognition dataset from the 2008 Pascal Large Scale Learning Challenge
- Using Searn for Entity Relation Recognition
- MNIST handwritten digit recognition with
- low-ranked quadratic
--lrqdemo on the movielens data-set
--normalizedeffect on SGD online learning
- Ability of
vwto separate signal from artificially injected noise
External sites examples
- Zygmunt Zając's FastML has several examples of using Vowpal Wabbit in practice.