Counterfactual Generative Networks

Abstract

Neural networks are prone to learning shortcuts -- they often model simple correlations, ignoring more complex ones that potentially generalize better. Prior works on image classification show that instead of learning a connection to object shape, deep classifiers tend to exploit spurious correlations with low-level texture or the background for solving the classification task. In this work, we take a step towards more robust and interpretable classifiers that explicitly expose the task's causal structure. Building on current advances in deep generative modeling, we propose to decompose the image generation process into independent causal mechanisms that we train without direct supervision. By exploiting appropriate inductive biases, these mechanisms disentangle object shape, object texture, and background; hence, they allow for generating counterfactual images. We demonstrate the ability of our model to generate such images on MNIST and ImageNet. Further, we show that the counterfactual images can improve out-of-distribution robustness with a marginal drop in performance on the original classification task, despite being synthetic. Lastly, our generative model can be trained efficiently on a single GPU, exploiting common pre-trained models as inductive biases.

Published at ICLR 2021

References in corpus (14)

• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Language Models are Few-Shot Learners
• U${}^2$-Net: Going Deeper with Nested U-Structure for Salient Object Detection
• Shortcut Learning in Deep Neural Networks
• ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
• Underspecification Presents Challenges for Credibility in Modern Machine Learning
• Invariant Risk Minimization
• Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World
• Causality for Machine Learning
• Explaining Classifiers with Causal Concept Effect (CaCE)
• Controlling generative models with continuous factors of variations
• Noise or Signal: The Role of Image Backgrounds in Object Recognition
• Selecting Data Augmentation for Simulating Interventions
• Towards causal generative scene models via competition of experts

Cited by in corpus (4)

• Finding and Fixing Spurious Patterns with Explanations
• Ensembling with Deep Generative Views
• Causal Learning for Socially Responsible AI
• Causal Discovery from Conditionally Stationary Time Series