Ranked batch-mode active learning
详细信息 查看全文
- 作者：Thiago N.C. Cardoso thiagon@dcc.ufmg.br" class="auth_mail" title="E-mail the corresponding author ; Rodrigo M. Silva rmsilva@dcc.ufmg.br" class="auth_mail" title="E-mail the corresponding author ; Sé ; rgio Canuto sergiodaniel@dcc.ufmg.br" class="auth_mail" title="E-mail the corresponding author ; Mirella M. Moro mirella@dcc.ufmg.br" class="auth_mail" title="E-mail the corresponding author ; Marcos A. Gonç ; alves ; mgoncalv@dcc.ufmg.br" class="auth_mail" title="E-mail the corresponding author
- 关键词：Active learning ; Batch-mode ; New ranking function ; Paradigm change
- 刊名：Information Sciences
- 出版年：2017
- 出版时间：10 February 2017
- 年：2017
- 卷：379
- 期：Complete
- 页码：313-337
- 全文大小：2962 K
文摘
- •
We introduce a new way of thinking about Batch-Mode Active Learning.
- •
A method for ranking unlabeled sets based on informativeness is proposed.
- •
Our results are superior (up to 25%) to pool-based batch-mode active learning.
- •
Our method is a drop-in replacement for batch-mode methods without their limitations.
- •
It is also superior to density-sensitive active learning methods.