• Publications
  • Influence
Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
TLDR
We show that even word embeddings trained on Google News articles exhibit female/male gender stereotypes to a disturbing extent. Expand
  • 1,069
  • 171
  • PDF
Adaptive Neural Networks for Efficient Inference
TLDR
We present an approach to adaptively utilize deep neural networks in order to reduce the evaluation time on new examples without loss of accuracy. Expand
  • 108
  • 7
  • PDF
The What-If Tool: Interactive Probing of Machine Learning Models
TLDR
We present the What-If Tool, an open-source application that allows practitioners to probe, visualize, and analyze ML systems, with minimal coding. Expand
  • 88
  • 6
  • PDF
XRAI: Better Attributions Through Regions
TLDR
We present a novel region-based attribution method, XRAI, that builds upon integrated gradients (Sundararajan et al. 2017), 2) introduce evaluation methods for empirically assessing the quality of image-based saliency maps, and 3) contribute an axiom-based sanity check for attribution methods. Expand
  • 27
  • 6
  • PDF
Adaptive Neural Networks for Fast Test-Time Prediction
TLDR
We present an approach to adaptively utilize deep neural networks in order to reduce the evaluation time on new examples without loss of classification performance. Expand
  • 24
  • 5
  • PDF
Quantifying and Reducing Stereotypes in Word Embeddings
TLDR
We study gender stereotypes in word embeddings trained only on word co-occurrence in text corpora, a popular framework to represent text data. Expand
  • 35
  • 5
  • PDF
BING++: A Fast High Quality Object Proposal Generator at 100fps
TLDR
We propose a novel object proposal algorithm {\em BING++} which inherits the good computational efficiency of BING \cite{BingObj2014} but significantly improves its proposal localization quality. Expand
  • 19
  • 2
  • PDF
The Language Interpretability Tool: Extensible, Interactive Visualizations and Analysis for NLP Models
TLDR
We present the Language Interpretability Tool (LIT), an open-source platform for visualization and understanding of NLP models. Expand
  • 18
  • 1
  • PDF
Debiasing Embeddings for Reduced Gender Bias in Text Classification
TLDR
We investigate the impact of gender bias in these pre-trained word embeddings on downstream classification tasks, using the case study of occupation classification. Expand
  • 11
  • 1
  • PDF
Structured Prediction with Test-time Budget Constraints
TLDR
We study the problem of structured prediction under test-time budget constraints, where the goal is to learn a system that is computationally efficient during test- time with little loss of predictive performance. Expand
  • 2
  • 1
  • PDF
...
1
2
3
...