no code implementations • 29 Sep 2021 • Kevin Alexander Laube, Maximus Mutschler, Andreas Zell
Due to inaccurate predictions, the selected architectures are generally suboptimal, which we quantify as an expected reduction in accuracy and hypervolume.
1 code implementation • 31 Aug 2021 • Maximus Mutschler, Kevin Laube, Andreas Zell
In the experiments conducted, our approach is on par with SGD with Momentum tuned with a piece-wise constant learning rate schedule and often outperforms other line search approaches for Deep Learning across models, datasets, and batch sizes on validation and test accuracy.
1 code implementation • 31 Mar 2021 • Maximus Mutschler, Andreas Zell
Optimization in Deep Learning is mainly guided by vague intuitions and strong assumptions, with a limited understanding how and why these work in practice.
no code implementations • 2 Oct 2020 • Maximus Mutschler, Andreas Zell
In traditional optimization, line searches are used to determine good step sizes, however, in deep learning, it is too costly to search for good step sizes on the expected empirical loss due to noisy losses.
1 code implementation • NeurIPS 2020 • Maximus Mutschler, Andreas Zell
The optimal step size is closely related to the shape of the loss in the update step direction.