Description:

This is a large scale online learning implementation with several useful features. See the webpage for more details.

Changes to previous version:

Initial Announcement on mloss.org.

BibTeX Entry: Download

Supported Operating Systems: Linux

Data Formats: None

Tags: Online Learning

Archive: download here

Comments

Ariel Faigon (on August 9, 2010, 06:33:42)

By far the fastest most scalable learning implementation I've used. Ideal for text analysis. On my desktop: generates a model from a 40,000 instances by 25 features in less than 2 seconds. - Supports highly sparse features, e.g. words and bi-grams in documents - Online algorithm (estimated error is test-like) - Very flexible input data format (features can be named, and/or partitioned into name-spaces), examples can be weighted and tagged. More friendly than SVMlight - Supports binary classification and regression - Automatic generation of quadratic features (crosses between given features) - Supports multiple loss functions (squared, hinge, logistic and quantile) The main limitation is that output is a simple weighted linear model and features should be monotonic with response to make sense.

Ariel Faigon (on August 9, 2010, 07:12:43)

Latest version is 4.1 and can be DL'ed from github: github.com/JohnLangford/vowpal_wabbit/downloads

Leave a comment

You must be logged in to post comments.