论文信息 - Component Random Walk

Component Random Walk

Label propagation has become a successful method for transductive learning. In this paper, we propose a unified label propagation model named Component Random Walk. We demonstrate that besides most of the existing label propagation algorithms, a novel Multilevel Component Propagation (MCP) algorithm can be derived from this Component Random Walk model as well. Promising experimental results are provided for MCP algorithm.

Ping He | Lin Lu | Ling Chen | Xiao-hua Xu | Zhou-Jin Pan

[1] Alexander Zien,et al. Semi-Supervised Classification by Low Density Separation , 2005, AISTATS.

[2] Jason Weston,et al. Structural bioinformatics Semi-supervised protein classification using cluster kernels , 2005 .

[3] Mikhail Belkin,et al. Manifold Regularization: A Geometric Framework for Learning from Labeled and Unlabeled Examples , 2006, J. Mach. Learn. Res..

[4] Arik Azran,et al. The rendezvous algorithm: multiclass semi-supervised learning with Markov random walks , 2007, ICML '07.

[5] Shih-Fu Chang,et al. Graph transduction via alternating minimization , 2008, ICML '08.

[6] Thorsten Joachims,et al. Transductive Inference for Text Classification using Support Vector Machines , 1999, ICML.

[7] Tommi S. Jaakkola,et al. Partially labeled classification with Markov random walks , 2001, NIPS.

[8] Thomas L. Griffiths,et al. Semi-Supervised Learning with Trees , 2003, NIPS.

[9] Jason Weston,et al. Semi-supervised Protein Classification Using Cluster Kernels , 2003, NIPS.

[10] Zoubin Ghahramani,et al. Combining active learning and semi-supervised learning using Gaussian fields and harmonic functions , 2003, ICML 2003.

[11] Bernhard Schölkopf,et al. Learning with Local and Global Consistency , 2003, NIPS.

[12] Thorsten Joachims,et al. Transductive Learning via Spectral Graph Partitioning , 2003, ICML.