Tell me who can learn you and I can tell you who you are: Landmarking Various Learning Algorithms

Landmarking is a novel approach to describing tasks in meta-learning. Previous approaches to meta-learning mostly considered only statistics-inspired measures of the data as a source for the de nition of metaattributes. Contrary to such approaches, landmarking tries to determine the location of a speci c learning problem in the space of all learning problems by directly measuring the performance of some simple and e cient learning algorithms themselves. In the experiments reported we show how such a use of landmark values can help to distinguish between areas of the learning space favouring di erent learners. Experiments, both with arti cial and real-world databases, show that landmarking selects, with moderate but reasonable level of success, the best performing of a set of learning algorithms.