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 1802.05074cs.LGstat.ML
收藏

L4:用于深度学习的实用的基于损失的步长自适应

L4: Practical loss-based stepsize adaptation for deep learning

Max-Planck-Institute for Intelligent Systems(马克斯·普朗克智能系统研究所)

Michal Rolinek, Georg Martius

摘要:

我们提出了一个步长自适应方案的随机梯度下降。它直接与损失函数一起操作,并重新调整梯度,以便对损失进行固定的预测进度。我们证明了它的能力,决定性地提高亚当和动量优化器的性能。具有默认超参数的增强型优化器始终优于其恒定步长的对应物,即使是最好的,也没有可测量的计算成本增加。性能在多种架构上进行了验证,包括密集网络,CNN,ResNets和经典数据集MNIST,时尚MNIST,CIFAR10等的递归差分神经计算机。

英文摘要:

We propose a stepsize adaptation scheme for stochastic gradient descent. It operates directly with the loss function and rescales the gradient in order to make fixed predicted progress on the loss. We demonstrate its capabilities by conclusively improving the performance of Adam and Momentum optimizers. The enhanced optimizers with default hyperparameters consistently outperform their constant stepsize counterparts, even the best ones, without a measurable increase in computational cost. The performance is validated on multiple architectures including dense nets, CNNs, ResNets, and the recurrent Differential Neural Computer on classical datasets MNIST, fashion MNIST, CIFAR10 and others.

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。

↑