ML Anthology

Feature Visualization

Chris Olah, Alexander Mordvintsev, Ludwig Schubert
Distill 2017
doi:10.23915/distill.00007 /distill/2017/olah2017distill-feature/

Abstract

Distill articles are interactive publications and do not include traditional abstracts. This summary was written for the ML Anthology. Presents a comprehensive overview of feature visualization techniques that reveal what neural networks learn by optimizing images to maximally activate specific neurons. Addresses practical challenges including high-frequency noise artifacts and introduces regularization and diversity methods.

Distill Code Semantic Scholar

Cite

Text

Olah et al. "Feature Visualization." Distill, 2017. doi:10.23915/distill.00007

Markdown

[Olah et al. "Feature Visualization." Distill, 2017.](https://mlanthology.org/distill/2017/olah2017distill-feature/) doi:10.23915/distill.00007

BibTeX

@article{olah2017distill-feature,
  title     = {{Feature Visualization}},
  author    = {Olah, Chris and Mordvintsev, Alexander and Schubert, Ludwig},
  journal   = {Distill},
  year      = {2017},
  doi       = {10.23915/distill.00007},
  url       = {https://mlanthology.org/distill/2017/olah2017distill-feature/}
}