Difference between revisions of "Daume and Marcu 2005 Learning as Search Optimization: Approximate Large Margin Methods for Structured Prediction"
From Cohen Courses
Jump to navigationJump to search| Line 11: | Line 11: | ||
=== Method === | === Method === | ||
| − | [[File: | + | [[File:LaSO Algorithm.png]] |
=== Experimental Result === | === Experimental Result === | ||
Revision as of 02:41, 1 October 2011
Citation and Online Link
Learning as Search Optimization: Approximate Large Margin Methods for Structured Prediction An alternative formal analysis of Searn.
Summary
The authors present the Learning as Search Optimization (LaSO) framework. The algorithm is basically SEARN but analyzed differently (and also ~24 pages shorter).
Like SEARN, LaSO attempts to combine the learning of the model with the search that occurs during decoding.
Method
Experimental Result
Related Papers
In progress by User:Jmflanig
