Continuation Methods for Mixing Heterogenous Sources
Adrian Corduneanu, Tommi Jaakkola
A number of modern learning tasks involve estimation from heterogeneous information sources. This includes classification with labeled and unlabeled data as well as other problems with analogous structure such as competitive (game theoretic) problems. The associated estimation problems can be typically reduced to solving a set of fixed point equations (consistency conditions). We introduce a general method for combining a preferred information source with another in this setting by evolving continuous paths of fixed points at intermediate allocations. We explicitly identify critical points along the unique paths to either increase the stability of estimation or to ensure a significant departure from the initial source. The homotopy continuation approach is guaranteed to terminate at the second source, and involves no combinatorial effort. We illustrate the power of these ideas both in classification tasks with labeled and unlabeled data, as well as in the context of a competitive (min-max) formulation of DNA sequence motif discovery.
Keywords: semi-supervised learning, partially-labeled data, EM, continuation
PS Link: http://www.ai.mit.edu/people/adrianc/publications/CorduneanuJaakkola02_UAI.ps.gz
PDF Link: /papers/02/p111-corduneanu.pdf
AUTHOR = "Adrian Corduneanu
and Tommi Jaakkola",
TITLE = "Continuation Methods for Mixing Heterogenous Sources",
BOOKTITLE = "Proceedings of the Eighteenth Conference Annual Conference on Uncertainty in Artificial Intelligence (UAI-02)",
PUBLISHER = "Morgan Kaufmann",
ADDRESS = "San Francisco, CA",
YEAR = "2002",
PAGES = "111--118"