arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2206.09034cs.LGcs.AIcs.CV
收藏

改进选择性分类

Towards Better Selective Classification

Mila – Université de Montréal(米拉-蒙特利尔大学);Borealis AI(北极光人工智能)

Leo Feng, Mohamed Osama Ahmed, Hossein Hajimirsadeghi, Amir Abdi

摘要:

我们解决选择性分类问题,其目标是在数据集的预定比例(覆盖率)上实现最佳性能。最近的最先进的选择方法通过引入单独的选择头或额外的选择logit来改变结构。在本文中,我们挑战上述方法。结果表明,最先进的方法的优越性能是由于训练一个更概括的分类器,而不是他们提出的选择机制。我们认为,表现最好的选择机制,而不是植根于分类器本身。我们提出的选择策略使用分类分数,并在所有覆盖范围和所有数据集上一致地实现更好的结果,而不会增加任何计算成本。此外,受半监督学习的启发,我们提出了一个基于熵的正则化,提高了选择性分类方法的性能。我们提出的选择机制与建议的基于熵的正则化实现了新的国家的最先进的结果。

英文摘要:

We tackle the problem of Selective Classification where the objective is to achieve the best performance on a predetermined ratio (coverage) of the dataset. Recent state-of-the-art selective methods come with architectural changes either via introducing a separate selection head or an extra abstention logit. In this paper, we challenge the aforementioned methods. The results suggest that the superior performance of state-of-the-art methods is owed to training a more generalizable classifier rather than their proposed selection mechanisms. We argue that the best performing selection mechanism should instead be rooted in the classifier itself. Our proposed selection strategy uses the classification scores and achieves better results by a significant margin, consistently, across all coverages and all datasets, without any added compute cost. Furthermore, inspired by semi-supervised learning, we propose an entropy-based regularizer that improves the performance of selective classification methods. Our proposed selection mechanism with the proposed entropy-based regularizer achieves new state-of-the-art results.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑