PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Transfer Learning With Adaptive Regularizers
Ulrich Rückert and Marius Kloft
In: Proceedings of the European Conference on Machine Learning (2011) Springer , Athens, Greece .

Abstract

The success of regularized risk minimization approaches to classication with linear models depends crucially on the selection of a regularization term that matches with the learning task at hand. If the necessary domain expertise is rare or hard to formalize, it may be di- cult to nd a good regularizer. On the other hand, if plenty of related or similar data is available, it is a natural approach to adjust the regularizer for the new learning problem based on the characteristics of the related data. In this paper, we study the problem of obtaining good parame- ter values for a `2-style regularizer with feature weights. We analytically investigate a moment-based method to obtain good values and give uni- form convergence bounds for the prediction error on the target learning task. An empirical study shows that the approach can improve predictive accuracy considerably in the application domain of text classicatio

PDF - Requires Adobe Acrobat Reader or other PDF viewer.
EPrint Type:Book Section
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
ID Code:9405
Deposited By:Marius Kloft
Deposited On:16 March 2012