arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2210.12499cs.CL
收藏

课程学习的培训动力:单语与跨语NLU研究

Training Dynamics for Curriculum Learning: A Study on Monolingual and Cross-lingual NLU

Huawei Noah’s Ark Lab(华为诺亚方舟实验室)

Fenia Christopoulou, Gerasimos Lampouras, Ignacio Iacobacci

摘要:

课程学习(Curriculum Learning,CL)是一种通过对样本进行排序来训练模型的技术,其目的是加速收敛和提高泛化能力。当前用于自然语言理解(NLU)任务的方法使用CL来通常通过面向对象或任务不可知的困难来提高分布中数据的性能。在这项工作中,相反,我们通过利用训练动态作为难度指标,即,统计数据,这些统计数据在训练期间测量特定任务数据实例上的模型行为,并基于这些统计数据提出对现有CL编译器的修改。与现有的工作不同,我们专注于在分布(ID),分布外(OOD)以及zero-shot(ZS)跨语言迁移数据集上评估模型。我们在几个NLU任务中表明,具有训练动态的CL可以在zero-shot跨语言迁移和OOD设置上产生更好的性能,在某些情况下可以提高8.5%。总的来说,实验表明,与其他难度指标相比,训练动态可以导致更好的模型性能,更平滑的训练,同时平均快20%。此外,通过分析,我们揭示了特定于任务的指标与与任务不可知的指标之间的相关性。

英文摘要:

Curriculum Learning (CL) is a technique of training models via ranking examples in a typically increasing difficulty trend with the aim of accelerating convergence and improving generalisability. Current approaches for Natural Language Understanding (NLU) tasks use CL to improve in-distribution data performance often via heuristic-oriented or task-agnostic difficulties. In this work, instead, we employ CL for NLU by taking advantage of training dynamics as difficulty metrics, i.e., statistics that measure the behavior of the model at hand on specific task-data instances during training and propose modifications of existing CL schedulers based on these statistics. Differently from existing works, we focus on evaluating models on in-distribution (ID), out-of-distribution (OOD) as well as zero-shot (ZS) cross-lingual transfer datasets. We show across several NLU tasks that CL with training dynamics can result in better performance mostly on zero-shot cross-lingual transfer and OOD settings with improvements up by 8.5% in certain cases. Overall, experiments indicate that training dynamics can lead to better performing models with smoother training compared to other difficulty metrics while being 20% faster on average. In addition, through analysis we shed light on the correlations of task-specific versus task-agnostic metrics.

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。

↑