Generative Interventions for Causal Learning
arXiv:2012.12265
Abstract
We introduce a framework for learning robust visual representations that generalize to new viewpoints, backgrounds, and scene contexts. Discriminative models often learn naturally occurring spurious correlations, which cause them to fail on images outside of the training distribution. In this paper, we show that we can steer generative models to manufacture interventions on features caused by confounding factors. Experiments, visualizations, and theoretical results show this method learns robust representations more consistent with the underlying causal relationships. Our approach improves performance on multiple datasets demanding out-of-distribution generalization, and we demonstrate state-of-the-art performance generalizing from ImageNet to ObjectNet dataset.
arXiv commentAccepted to CVPR 2021
References in corpus (12)
- Auto-Encoding Variational Bayes
- Large Scale GAN Training for High Fidelity Natural Image Synthesis
- Improved Regularization of Convolutional Neural Networks with Cutout
- The Effectiveness of Data Augmentation in Image Classification using Deep Learning
- Data Augmentation Generative Adversarial Networks
- ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
- Causality for Machine Learning
- CXPlain: Causal Explanations for Model Interpretation under Uncertainty
- Causal Induction from Visual Observations for Goal Directed Tasks
- Selecting Data Augmentation for Simulating Interventions
- Continuously Indexed Domain Adaptation
- Deep causal representation learning for unsupervised domain adaptation