Abstract
Landmarking is a recent and promising meta-learning strategy, which defines meta-features that are themselves efficient learning algorithms. However, the choice of landmarkers is often made in an ad hoc manner. In this paper, we propose a new perspective and set of criteria for landmarkers. Based on the new criteria, we propose a landmarker generation algorithm, which generates a set of landmarkers that are each subsets of the algorithms being landmarked. Our experiments show that the landmarkers formed, when used with linear regression are able to estimate the accuracy of a set of candidate algorithms well, while only utilising a small fraction of the computational cost required to evaluate those candidate algorithms via ten-fold cross-validation.
Original language | English |
---|---|
Pages (from-to) | 296-306 |
Number of pages | 11 |
Journal | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
Volume | 3339 |
DOIs | |
Publication status | Published - 2004 |
Externally published | Yes |
Event | 17th Australian Joint Conference on Artificial Intelligence, AI 2004: Advances in Artificial Intelligence - Cairns, Australia Duration: 4 Dec 2004 → 6 Dec 2004 |