Transfer Learning, Soft Distance-Based Bias, and the Hierarchical BOA
This work addresses efficiency improvements for practitioners using hBOA on complex optimization problems, though it appears incremental as it builds on existing transfer learning methods.
The paper tests a transfer learning technique for the hierarchical Bayesian optimization algorithm (hBOA) on NP-complete problems like MAXSAT, showing it improves efficiency even across different problem sizes and yields nearly multiplicative speedups when combined with other methods.
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.