The design of fast algorithms stands out as a clear example of eective incremental
research: there are clear chains of advances in fast rule learning and in
fast decision-tree learning. Research on partitioned-data approaches and on mining
relationally represented data is less mature, consisting mainly of independent
work, but there are striking commonalities among the existing approaches. We
believe these areas are ready for more emphasis on comparative research, followed
by (hopefully signicant) incremental advances.