ABSTRACT
These new characteristics of big data have brought new challenges and opportunities for classification algorithms. For example, to address the challenge of velocity, on-line streaming classification algorithms have been proposed (please refer to the previous chapter for details); in order to address the challenge of variety, the so-called heterogeneous machine learning has been emerging, including multi-view classification for data heterogeneity, transfer learning and multi-task classification for classification task heterogeneity, multi-instance learning for instance heterogeneity, and
classification with crowd-sourcing for oracle heterogeneity; in order to address the challenge of volume, many efforts have been devoted to scaling up classification algorithms.
