Lazy Transformation-Based Learning

Weintroduce a significant improvementfor a relatively newmachine learning methodcalled Transformation-Based Learning. By applying a MonteCarlo strategy to randomly sample from the space of rules, rather than exhaustively analyzing all possible rules, wedrastically reducethe memoryand time costs of the algorithm, without compromisingaccuracy on unseen data. This enables Transformation-BasedLearning to apply to a widerrange of domains,as it can effectively consider a larger numberof different features and feature interactions in the data. In addition, the MonteCarlo improvementdecreases the labor demands on the humandeveloper, who no longer needs to develop a minimalset of rule templates to maintaintractability.

• Type: Conference
• Year: 1998
• Where: FLAIRS
• Authors: Ken Samuel
• Keywords: Artificial Intelligence, FLAIRS 1998, Memoryand Time Costs, MonteCarlo Strategy, Transformation-Based Learning