Active learning based on minimization of the expected path-length of random walks on the learned manifold structure

作者：

Highlights：

• Active learning based on the expected path-length of random walks is feasible.

• The selected query samples prove to have a guaranteed solution bound.

• An approach to tuning of the RBF kernel parameter for active learning is proposed.

摘要

•Active learning based on the expected path-length of random walks is feasible.•The selected query samples prove to have a guaranteed solution bound.•An approach to tuning of the RBF kernel parameter for active learning is proposed.

论文关键词：Active learning,Locally linear embedding,Random walks,Submodular set functions

论文评审过程：Received 20 January 2017, Revised 8 May 2017, Accepted 1 June 2017, Available online 6 June 2017, Version of Record 22 June 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.06.001