Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV) ​

← Back to topic

Authors: Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, Rory Sayres
Year: 2017
Journal: ICML
DOI: 10.48550/arXiv.1711.11279
Publisher: https://arxiv.org/abs/1711.11279

Keywords: tcav, concept

Abstract ​

We introduce TCAV a method for interpreting neural networks using high-level concepts.

Cite this paper ​

@misc{tcav2017,
  title  = {Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)},
  author = {Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, Rory Sayres},
  year   = {2017},
  journal = {ICML},
  doi    = {10.48550/arXiv.1711.11279},
  url    = {https://doi.org/10.48550/arXiv.1711.11279},
}

Source files ​

• metadata.json
• paper.bib