Deep transfer via second-order Markov logic

Jesse Davis, Pedro Domingos

Abstract

Standard inductive learning requires that training and test instances come from the same distribution. Transfer learning seeks to remove this restriction. In shallow trans-fer, test instances are from the same do-main, but have a different distribution. In deep transfer, test instances are from a dif-ferent domain entirely (i.e., described by dif-ferent predicates). Humans routinely per-form deep transfer, but few learning sys-tems, if any, are capable of it. In this pa-per we propose an approach based on a form of second-order Markov logic. Our algorithm discovers structural regularities in the source domain in the form of Markov logic for-mulas with predicate variables, and instan-tiates these formulas with predicates from the target domain. Using this approach, we have successfully transferred learned knowl-edge among molecular biology, social network and Web domains. The discovered patterns include broadly useful properties of predi-cates, like symmetry and transitivity, and relations among predicates, such as various forms of homophily. 1.
BibTeX
@inproceedings{davis2009deep,
  author = {Jesse Davis and Pedro Domingos},
  title = {{Deep transfer via second-order Markov logic}},
  booktitle = {International Conference on Machine Learning},
  year = {2009},
  doi = {10.1145/1553374.1553402},
}

Loading related content...

Figure