Backforward Propagation (Student Abstract)

AAAI 2023 pp. 16338-16339

doi:10.1609/AAAI.V37I13.27029 /aaai/2023/stoica2023aaai-backforward/

Abstract

In this paper we introduce Backforward Propagation, a method of completely eliminating Internal Covariate Shift (ICS). Unlike previous methods, which only indirectly reduce the impact of ICS while introducing other biases, we are able to have a surgical view at the effects ICS has on training neural networks. Our experiments show that ICS has a weight regularizing effect on models, and completely removing it enables for faster convergence of the neural network.

PDF AAAI Semantic Scholar

Cite

Text

Stoica and Simionescu. "Backforward Propagation (Student Abstract)." AAAI Conference on Artificial Intelligence, 2023. doi:10.1609/AAAI.V37I13.27029

Markdown

[Stoica and Simionescu. "Backforward Propagation (Student Abstract)." AAAI Conference on Artificial Intelligence, 2023.](https://mlanthology.org/aaai/2023/stoica2023aaai-backforward/) doi:10.1609/AAAI.V37I13.27029

BibTeX

@inproceedings{stoica2023aaai-backforward,
  title     = {{Backforward Propagation (Student Abstract)}},
  author    = {Stoica, George and Simionescu, Cristian},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2023},
  pages     = {16338-16339},
  doi       = {10.1609/AAAI.V37I13.27029},
  url       = {https://mlanthology.org/aaai/2023/stoica2023aaai-backforward/}
}