• Corpus ID: 53467348

Feature-Wise Bias Amplification

  title={Feature-Wise Bias Amplification},
  author={Klas Leino and Matt Fredrikson and Emily Black and Shayak Sen and Anupam Datta},
We study the phenomenon of bias amplification in classifiers, wherein a machine learning model learns to predict classes with a greater disparity than the underlying ground truth. We demonstrate that bias amplification can arise via an inductive bias in gradient descent methods that results in the overestimation of the importance of moderately-predictive "weak" features if insufficient training data is available. This overestimation gives rise to feature-wise bias amplification -- a previously… 

Figures and Tables from this paper

A Systematic Study of Bias Amplification

Recent research suggests that predictions made by machine-learning models can amplify biases present in the training data. When a model amplifies bias, it makes certain predictions at a higher rate

Directional Bias Amplification

A new, decoupled metric for measuring bias amplification, BiasAmp→ (Directional Bias Amplification), is introduced and analyzed and thoroughly analyze and discusses both the technical assumptions and the normative implications of this metric.

Gradient Based Activations for Accurate Bias-Free Learning

This work shows that a biased discriminator can actually be used to improve this bias-accuracy tradeoff by using a feature masking approach using the discriminator's gradients, and shows that this simple approach works well to reduce bias as well as improve accuracy significantly.

Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigation

A simple but surprisingly effective visual recognition benchmark for studying bias mitigation, and a simple but similarly effective alternative to the inference-time Reducing Bias Amplification method of Zhao et al., and design a domain-independent training technique that outperforms all other methods.

Data Feedback Loops: Model-driven Amplification of Dataset Biases

Experiments in three conditional prediction scenarios demonstrate that models that exhibit a sampling-like behavior are more calibrated and thus more stable, and an intervention is proposed to help calibrate and stabilize unstable feedback systems.

Men Also Do Laundry: Multi-Attribute Bias Amplification

This work proposes a new metric: Multi-Attribute Bias Amplification, which refers to models amplifying inherent training set biases at test time and validate the proposed metric through an anal- ysis of gender bias ampliﷁcation on the COCO and imSitu datasets.

Topics to Avoid: Demoting Latent Confounds in Text Classification

This work proposes a method that represents the latent topical confounds and a model which “unlearns” confounding features by predicting both the label of the input text and the confound; but it shows that this model generalizes better and learns features that are indicative of the writing style rather than the content.

A Modulation Layer to Increase Neural Network Robustness Against Data Quality Issues

A novel neural network modification to mitigate the impacts of low quality and missing data which involves replacing the weights of a fully-connected layer with a function of an additional input based on their reliability and the presence of other data.

Data-Centric Factors in Algorithmic Fairness

A new dataset on recidivism in 1.5 million criminal cases from courts in the U.S. state of Wisconsin, 2000-2018 is introduced and it is found that factors often do influence fairness metrics holding the classifier specification constant, without having a corresponding effect on accuracy metrics.

Measuring Biases of Word Embeddings: What Similarity Measures and Descriptive Statistics to Use?

It is found that over the ten categories of word embedding association tests, Mahalanobis distance reveals the smallest bias, and Euclidean Distance reveals the largest bias in word embeddings.



ConvNets and ImageNet Beyond Accuracy: Explanations, Bias Detection, Adversarial Examples and Model Criticism

It is shown that explanations can mitigate the impact of misclassified adversarial examples from the perspective of the end-user and a novel tool for uncovering the undesirable biases learned by a model is introduced.

A systematic study of the class imbalance problem in convolutional neural networks

The Implicit Bias of Gradient Descent on Separable Data

We examine gradient descent on unregularized logistic regression problems, with homogeneous linear predictors on linearly separable datasets. We show the predictor converges to the direction of the

Learning Multiple Layers of Features from Tiny Images

It is shown how to train a multi-layer generative model that learns to extract meaningful features which resemble those found in the human visual cortex, using a novel parallelization algorithm to distribute the work among multiple machines connected on a network.

Class Imbalance, Redux

This work approaches the problem of imbalance from a probabilistic perspective, and from this vantage identify dataset characteristics that exacerbate the problem and advocates the approach of bagging an ensemble of classifiers induced over balanced bootstrap training samples, arguing that this strategy will often succeed where others fail.

Sampling Bias and Class Imbalance in Maximum-likelihood Logistic Regression

Logistic regression is a widely used statistical method to relate a binary response variable to a set of explanatory variables and maximum likelihood is the most commonly used method for parameter

Women also Snowboard: Overcoming Bias in Captioning Models

A new Equalizer model is introduced that ensures equal gender probability when gender Evidence is occluded in a scene and confident predictions when gender evidence is present and has lower error than prior work when describing images with people and mentioning their gender and more closely matches the ground truth ratio of sentences including women to sentences including men.

Neural Networks and the Bias/Variance Dilemma

It is suggested that current-generation feedforward neural networks are largely inadequate for difficult problems in machine perception and machine learning, regardless of parallel-versus-serial hardware or other implementation issues.

Deep Learning Face Attributes in the Wild

A novel deep learning framework for attribute prediction in the wild that cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently.

Influence-Directed Explanations for Deep Convolutional Networks

Evaluation demonstrates that influence-directed explanations identify influential concepts that generalize across instances, can be used to extract the “essence” of what the network learned about a class, and isolate individual features the network uses to make decisions and distinguish related classes.