Visualizing Memorization in RNNs

Abstract

Distill articles are interactive publications and do not include traditional abstracts. This summary was written for the ML Anthology. Introduces a connectivity visualization technique that maps gradient magnitudes between inputs and outputs of recurrent neural networks, revealing differences in how GRU, LSTM, and Nested LSTM architectures use short-term versus long-term context.

Cite

Text

Madsen. "Visualizing Memorization in RNNs." Distill, 2019. doi:10.23915/distill.00016

Markdown

[Madsen. "Visualizing Memorization in RNNs." Distill, 2019.](https://mlanthology.org/distill/2019/madsen2019distill-visualizing/) doi:10.23915/distill.00016

BibTeX

@article{madsen2019distill-visualizing,
  title     = {{Visualizing Memorization in RNNs}},
  author    = {Madsen, Andreas},
  journal   = {Distill},
  year      = {2019},
  doi       = {10.23915/distill.00016},
  url       = {https://mlanthology.org/distill/2019/madsen2019distill-visualizing/}
}