Looking at Deep Learning Phenomena Through a Telescoping Lens

Abstract

Deep learning sometimes appears to work in unexpected ways. In pursuit of deeper understanding of its surprising behaviors, we investigate the utility of a tractable and accurate model of a neural network consisting of a sequence of first-order approximations _telescoping_ out into a single empirically operational tool for practical analysis. We illustrate how it can be applied to derive new empirical insights on a diverse range of prominent phenomena in the literature -- including double descent, grokking, and the challenges of applying deep learning on tabular data.

Cite

Text

Jeffares et al. "Looking at Deep Learning Phenomena Through a Telescoping Lens." ICML 2024 Workshops: HiLD, 2024.

Markdown

[Jeffares et al. "Looking at Deep Learning Phenomena Through a Telescoping Lens." ICML 2024 Workshops: HiLD, 2024.](https://mlanthology.org/icmlw/2024/jeffares2024icmlw-looking/)

BibTeX

@inproceedings{jeffares2024icmlw-looking,
  title     = {{Looking at Deep Learning Phenomena Through a Telescoping Lens}},
  author    = {Jeffares, Alan and Curth, Alicia and van der Schaar, Mihaela},
  booktitle = {ICML 2024 Workshops: HiLD},
  year      = {2024},
  url       = {https://mlanthology.org/icmlw/2024/jeffares2024icmlw-looking/}
}