Skip to content

Latest commit

 

History

History
25 lines (18 loc) · 831 Bytes

KohLia17.md

File metadata and controls

25 lines (18 loc) · 831 Bytes

Understanding Black-box Predictions via Influence Functions

Pang Wei Koh, Percy Liang. Understanding Black-Box Predictions via Influence Functions. ICML 2017.

tl;dr

  • Takeaway 1
  • Takeaway 2
  • Open question or critique

Contributions

Using influence functions, the authors can trace a prediction back to the training data.

Previous work

  • TODO

Influence Functions

  • The classical result allows you to determine the influence of upweighting training data point on parameters theta. We can use the
  • What are the main contributions?
  • Point to relevant long parts of paper that are unable to be summarized

Applications

  • Image classification

Q's for authors

TODO