A Review of Sequential Supervised Learning

Abstract

Sequential supervised learning problems arise in many applications. After a definition of this learning task, we give out a set of evaluation criteria to the sequential supervised learning algorithms, some leading ones of which are described in this paper and evaluated based on the given criteria. In this paper, we show how these sequential supervised learning algorithms evolve from one to another based on the analysis of their limitations and advantages. We also try to give some intuitions to design novel sequential learning algorithms by showing how the existing ones are motivated by the corresponding traditional supervised learning algorithms. Some open problems in this area are discussed at the end