arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2108.07686cs.LG
收藏

深度学习的缩放定律

Scaling Laws for Deep Learning

Israel Institute of Technology(以色列理工学院);Massachusetts Institute of Technology(麻省理工学院)

Jonathan S. Rosenfeld

摘要:

跑得更快只能让你走这么远--通常建议首先了解道路通向哪里,然后再开车。在过去十年中,机器学习(ML)和深度学习(DL)的复兴伴随着不可扩展的计算成本,限制了其进步并在实践中对该领域产生影响。在这篇论文中,我们采取系统的方法来解决这些成本的根源算法和方法的局限性。我们首先证明了DL训练和修剪是可预测的,并由缩放定律控制-对于最先进的模型和任务,跨越图像分类和语言建模,以及通过迭代修剪进行最先进的模型压缩。可预测性,通过建立这些缩放定律,提供了原则性的设计和权衡推理,目前在很大程度上缺乏在该领域的路径。然后,我们继续分析来源的缩放定律,提供一个近似理论的观点,并显示通过探索一个无噪声的可实现的情况下,DL实际上是占主导地位的误差源非常远离误差下限。最后,我们建立在缩放定律的起源的理论理解。我们提出了一种方法来消除当前的主要误差源之一-通过数据带宽限制假设和奈奎斯特学习器的引入-这可以在原则上达到泛化误差下限(例如,在无噪声的情况下为0),在有限的数据集大小。

英文摘要:

Running faster will only get you so far -- it is generally advisable to first understand where the roads lead, then get a car ... The renaissance of machine learning (ML) and deep learning (DL) over the last decade is accompanied by an unscalable computational cost, limiting its advancement and weighing on the field in practice. In this thesis we take a systematic approach to address the algorithmic and methodological limitations at the root of these costs. We first demonstrate that DL training and pruning are predictable and governed by scaling laws -- for state of the art models and tasks, spanning image classification and language modeling, as well as for state of the art model compression via iterative pruning. Predictability, via the establishment of these scaling laws, provides the path for principled design and trade-off reasoning, currently largely lacking in the field. We then continue to analyze the sources of the scaling laws, offering an approximation-theoretic view and showing through the exploration of a noiseless realizable case that DL is in fact dominated by error sources very far from the lower error limit. We conclude by building on the gained theoretical understanding of the scaling laws' origins. We present a conjectural path to eliminate one of the current dominant error sources -- through a data bandwidth limiting hypothesis and the introduction of Nyquist learners -- which can, in principle, reach the generalization error lower limit (e.g. 0 in the noiseless case), at finite dataset size.

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。

↑