arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2210.03005cs.LGcs.AIcs.CLcs.DB
收藏

要不要 Softmax,这是个问题:在 Transformer 模型上应用主动学习时的选择

To Softmax, or not to Softmax: that is the question when applying Active Learning for Transformer Models

Technische Universität Dresden(德累斯顿工业大学);Hochschule für Technik und Wirtschaft Dresden(德累斯顿工程与经济应用科学大学)

Julius Gonsior, Christian Falkenberg, Silvio Magino, Anja Reusch, Maik Thiele, Wolfgang Lehner

AI总结:

本文研究 Transformer 模型主动学习中 softmax 置信度不可靠的问题,在七个数据集上比较八种替代方法,发现只标注离群点会降低性能,并提出系统性忽略离群样本的启发式策略以改进多种主动学习方法。

摘要:

尽管基于 Transformer 的语言模型在几乎所有自然语言处理应用中取得了最先进的结果,但对其进行微调仍然需要大量标注数据才能有效工作。一种众所周知的可减少获取标注数据集所需人力的技术是主动学习(Active Learning, AL):这是一个迭代过程,过程中只对最小数量的样本进行标注。主动学习策略需要获得模型预测的量化置信度度量。一个常见的选择是在最后一层使用 softmax 激活函数。由于 softmax 函数会提供具有误导性的概率,本文在七个数据集上比较了八种替代方法。我们几乎自相矛盾的发现是,大多数方法过于擅长识别真正最不确定的样本(离群点),因此只标注离群点反而会导致性能变差。作为一种启发式策略,我们提出系统地忽略这些样本,与 softmax 函数相比,这一做法使多种方法取得了改进。

英文摘要:

Despite achieving state-of-the-art results in nearly all Natural Language Processing applications, fine-tuning Transformer-based language models still requires a significant amount of labeled data to work. A well known technique to reduce the amount of human effort in acquiring a labeled dataset is \textit{Active Learning} (AL): an iterative process in which only the minimal amount of samples is labeled. AL strategies require access to a quantified confidence measure of the model predictions. A common choice is the softmax activation function for the final layer. As the softmax function provides misleading probabilities, this paper compares eight alternatives on seven datasets. Our almost paradoxical finding is that most of the methods are too good at identifying the true most uncertain samples (outliers), and that labeling therefore exclusively outliers results in worse performance. As a heuristic we propose to systematically ignore samples, which results in improvements of various methods compared to the softmax function.

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。

↑