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 1609.04836cs.LGmath.OC
收藏

关于深度学习的大批量训练:泛化间隙和尖锐极小值

On Large-Batch Training for Deep Learning: Generalization Gap and Sharp Minima

Northwestern University(西北大学);Intel Corporation(英特尔公司)

Nitish Shirish Keskar, Dheevatsa Mudigere, Jorge Nocedal, Mikhail Smelyanskiy, Ping Tak Peter Tang

摘要:

随机梯度下降(SGD)方法及其变体是许多深度学习任务的首选算法。这些方法以小批量方式操作,其中对训练数据的一部分(例如$32$ - $512$数据点)进行采样以计算对梯度的近似。在实践中已经观察到,当使用较大的批次时,模型的质量会下降,如通过其概括能力所测量的。我们调查的原因,这种泛化下降的大批量政权和目前的数值证据支持的观点,即大批量的方法往往收敛到尖锐的极小值的训练和测试功能-众所周知,尖锐的极小值导致较差的泛化。相比之下,小批量方法始终收敛到平坦的最小值,我们的实验支持一个普遍持有的观点,这是由于梯度估计中的固有噪声。我们讨论了几种策略,试图帮助大批量方法消除这种泛化差距。

英文摘要:

The stochastic gradient descent (SGD) method and its variants are algorithms of choice for many Deep Learning tasks. These methods operate in a small-batch regime wherein a fraction of the training data, say $32$-$512$ data points, is sampled to compute an approximation to the gradient. It has been observed in practice that when using a larger batch there is a degradation in the quality of the model, as measured by its ability to generalize. We investigate the cause for this generalization drop in the large-batch regime and present numerical evidence that supports the view that large-batch methods tend to converge to sharp minimizers of the training and testing functions - and as is well known, sharp minima lead to poorer generalization. In contrast, small-batch methods consistently converge to flat minimizers, and our experiments support a commonly held view that this is due to the inherent noise in the gradient estimation. We discuss several strategies to attempt to help large-batch methods eliminate this generalization gap.

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。

↑