Missingness Bias in Model Debugging

2022/04/19 by Jain, Saachi, Salman, Hadi, Wong, Eric +4 · 5 citations
#Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2204.08945

Abstract

Missingness, or the absence of features from an input, is a concept fundamental to many model debugging tools. However, in computer vision, pixels cannot simply be removed from an image. One thus tends to resort to heuristics such as blacking out pixels, which may in turn introduce bias into the debugging process. We study such biases and, in particular, show how transformer-based architectures can enable a more natural implementation of missingness, which side-steps these issues and improves the reliability of model debugging in practice. Our code is available at https://github.com/madrylab/missingness

Cited by

• Back to the Baseline: Examining Baseline Effects on Explainability Metrics
• Concept Regions Matter: Benchmarking CLIP with a New Cluster-Importance Approach
• Improving Perturbation-based Explanations by Understanding the Role of Uncertainty Calibration
• Rethinking Explanation Evaluation under the Retraining Scheme
• Detecting Regional Spurious Correlations in Vision Transformers via Token Discarding

Related

• MissingBench-Verified: Probing Vision-Language Models' Inability to Detect Missing Object Parts
• VISION DIFFMASK: Faithful Interpretation of Vision Transformers with Differentiable Patch Masking
• HiBug2: Efficient and Interpretable Error Slice Discovery for Comprehensive Model Debugging
• Improved anomaly detection by training an autoencoder with skip connections on images corrupted with Stain-shaped noise
• Exploring Corruption Robustness: Inductive Biases in Vision Transformers and MLP-Mixers