 |
    |
Transfer Learning, Soft Distance-Based Bias, and the Hierarchical BOA
Martin Pelikan1, Mark W. Hauschild1, and Pier Luca Lanzi2
1Missouri Estimation of Distribution Algorithms Laboratory (MEDAL), Department of Mathematics and Computer Science, University of Missouri, St. Louis, MO 63121 USA
martin@martinpelikan.net
mwh308@umsl.edu
http://medal-lab.org/
2Dipartimento di Elettronica e Informazione, Politecnico di Milano, Piazza Leonardo da Vinci, 32, I-20133, Milano, Italy
pierluca.lanzi@polimi.it
http://www.pierlucalanzi.net/
Abstract. An automated technique has recently been proposed to transfer learning in the hierarchical Bayesian optimization algorithm (hBOA) based on distance-based statistics. The technique enables practitioners to improve hBOA efficiency by collecting statistics from probabilistic models obtained in previous hBOA runs and using the obtained statistics to bias future hBOA runs on similar problems. The purpose of this paper is threefold: (1) test the technique on several classes of NP-complete problems, including MAXSAT, spin glasses and minimum vertex cover; (2) demonstrate that the technique is effective even when previous runs were done on problems of different size; (3) provide empirical evidence that combining transfer learning with other efficiency enhancement techniques can often yield nearly multiplicative speedups.
Keywords: Transfer learning, inductive transfer, learning from experience, estimation of distribution algorithms, hierarchical Bayesian optimization algorithm, decomposable problems, efficiency enhancement
LNCS 7491, p. 173 ff.
lncs@springer.com
© Springer-Verlag Berlin Heidelberg 2012