Adaptive graph Laplacian MTL L1, L2 and LS-SVMs
- 22 March 2024
- journal article
- research article
- Published by Oxford University Press (OUP) in Logic Journal of the IGPL
- Vol. 32 (4) , 634-655
Abstract
Multi-Task Learning tries to improve the learning process of different tasks by solving them simultaneously. A popular Multi-Task Learning formulation for SVM is to combine common and task-specific parts. Other approaches rely on using a Graph Laplacian regularizer. Here we propose a combination of these two approaches that can be applied to L1, L2 and LS-SVMs. We also propose an algorithm to iteratively learn the graph adjacency matrix used in the Laplacian regularization. We test our proposal with synthetic and real problems, both in regression and classification settings. When the task structure is present, we show that our model is able to detect it, which leads to better results, and we also show it to be competitive even when this structure is not present.Keywords
This publication has 15 references indexed in Scilit:
- Convex formulation for multi-task L1-, L2-, and LS-SVMsNeurocomputing, 2021
- Learning output kernels for multi-task problemsNeurocomputing, 2013
- Simple Proof of Convergence of the SMO Algorithm for Different SVM VariantsIEEE Transactions on Neural Networks and Learning Systems, 2012
- Generalized SMO Algorithm for SVM-Based Multitask LearningIEEE Transactions on Neural Networks and Learning Systems, 2012
- Convex multi-task feature learningMachine Learning, 2008
- Learning to learn with the informative vector machine2004
- A Generalized Representer Theorem2001
- The Nature of Statistical Learning Theory2000
- Multitask LearningMachine Learning, 1997
- Hierarchical Bayes Conjoint Analysis: Recovery of Partworth Heterogeneity from Reduced Experimental DesignsMarketing Science, 1996