论文标题
通过虚拟对抗扰动对不确定性句子进行抽样
Uncertainty Sentence Sampling by Virtual Adversarial Perturbation
论文作者
论文摘要
通过确定最有用的示例来降低注释成本的主动学习试图降低注释成本。在基于池的场景中，主动学习的常用方法使用不确定性或多样性采样。在这项工作中，为了纳入预测性不确定性和样本多样性，我们提出了使用虚拟对抗性扰动（Miyato等，2019）作为模型表示，使用虚拟对抗性扰动（Miyato等人，2019年），将虚拟的对抗性扰动（Vapal）（Vapal）（Vapal），一种不确定性多样性组合框架。 Vapal在四个句子理解数据集上的表现始终如一，甚至比强大的基线：Agnews，IMDB，PubMed和SST-2更好，为在句子理解任务上的句子理解方面提供了潜在的选择。
Active learning for sentence understanding attempts to reduce the annotation cost by identifying the most informative examples. Common methods for active learning use either uncertainty or diversity sampling in the pool-based scenario. In this work, to incorporate both predictive uncertainty and sample diversity, we propose Virtual Adversarial Perturbation for Active Learning (VAPAL) , an uncertainty-diversity combination framework, using virtual adversarial perturbation (Miyato et al., 2019) as model uncertainty representation. VAPAL consistently performs equally well or even better than the strong baselines on four sentence understanding datasets: AGNEWS, IMDB, PUBMED, and SST-2, offering a potential option for active learning on sentence understanding tasks.
