Separability versus prototypicality in handwritten word-image retrieval

作者：

Highlights：

• Good classifiers do not intrinsically imply a proper ranking of instances.

• A feature that is good for classification may not be optimal for ranking.

• Retrieval from massive data sets requires an optimal reduction of candidates.

• A reduction of candidates requires good separability.

• At the same time, proper ranking requires an intuitive prototypicality of instances.

摘要

Highlights•Good classifiers do not intrinsically imply a proper ranking of instances.•A feature that is good for classification may not be optimal for ranking.•Retrieval from massive data sets requires an optimal reduction of candidates.•A reduction of candidates requires good separability.•At the same time, proper ranking requires an intuitive prototypicality of instances.

论文关键词：Image retrieval,Handwriting recognition,Nearest centroid,Support-vector machines,Separability,Prototypicality,Historical manuscripts,Big data,Continuous machine learning

论文评审过程：Available online 20 September 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.09.006