Semi-supervised Batch Active Learning via Bilevel Optimization

2020/10/19 by Zalán Borsos, Marco Tagliasacchi, Borsos, Zalán +3
Computer Science · Mathematics · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Sound (cs.SD) #cs.LG #cs.SD #stat.ML

paper · pdf · doi:10.48550/arxiv.2010.09654

arxiv created 2020/10/19 · arxiv updated 2020/10/20

Abstract

Active learning is an effective technique for reducing the labeling cost by improving data efficiency. In this work, we propose a novel batch acquisition strategy for active learning in the setting where the model training is performed in a semi-supervised manner. We formulate our approach as a data summarization problem via bilevel optimization, where the queried batch consists of the points that best summarize the unlabeled data pool. We show that our method is highly effective in keyword detection tasks in the regime when only few labeled samples are available.

Related

• Learning to Rank for Active Learning via Multi-Task Bilevel Optimization
• Diverse mini-batch Active Learning
• Batch Active Learning from the Perspective of Sparse Approximation
• Batch Active Learning at Scale
• Bayesian Batch Active Learning as Sparse Subset Approximation