# Measuring the tendency of CNNs to Learn Surface Statistical Regularities

arXiv: https://arxiv.org/abs/1711.11561 · HTML: https://mlsanity.com/paper/1711.11561/

## Key facts

- Cited by 87 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 30 Nov 2017.

## Built on by (most cited first, 30 of 87)

- [Vision Transformers are Robust Learners](https://mlsanity.com/paper/2105.07581/) (2021)
- [Partial success in closing the gap between human and machine vision](https://mlsanity.com/paper/2106.07411/) (2021)
- [3D Common Corruptions and Data Augmentation](https://mlsanity.com/paper/2203.01441/) (2022)
- [Rethinking Architecture Design for Tackling Data Heterogeneity in Federated Learning](https://mlsanity.com/paper/2106.06047/) (2021)
- [Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU Models](https://mlsanity.com/paper/2103.06922/) (2021)
- [On the Foundations of Shortcut Learning](https://mlsanity.com/paper/2310.16228/) (2023)
- [MulT: An End-to-End Multitask Learning Transformer](https://mlsanity.com/paper/2205.08303/) (2022)
- [A Survey on Transferability of Adversarial Examples across Deep Neural Networks](https://mlsanity.com/paper/2310.17626/) (2023)
- [Robust Contrastive Learning Using Negative Samples with Diminished Semantics](https://mlsanity.com/paper/2110.14189/) (2021)
- [Spectral Bias in Practice: The Role of Function Frequency in Generalization](https://mlsanity.com/paper/2110.02424/) (2021)
- [A Little Robustness Goes a Long Way: Leveraging Robust Features for Targeted Transfer Attacks](https://mlsanity.com/paper/2106.02105/) (2021)
- [Does enhanced shape bias improve neural network robustness to common corruptions?](https://mlsanity.com/paper/2104.09789/) (2021)
- [Downstream-agnostic Adversarial Examples](https://mlsanity.com/paper/2307.12280/) (2023)
- [Exploring the Efficacy of Automatically Generated Counterfactuals for Sentiment Analysis](https://mlsanity.com/paper/2106.15231/) (2021)
- [Towards Trustworthy and Aligned Machine Learning: A Data-centric Survey with Causality Perspectives](https://mlsanity.com/paper/2307.16851/) (2023)
- [Don't blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropy](https://mlsanity.com/paper/2308.12553/) (2023)
- [Test time Adaptation through Perturbation Robustness](https://mlsanity.com/paper/2110.10232/) (2021)
- [Domain Generalization by Learning and Removing Domain-specific Features](https://mlsanity.com/paper/2212.07101/) (2022)
- [What do neural networks learn in image classification? A frequency shortcut perspective](https://mlsanity.com/paper/2307.09829/) (2023)
- [Dynamic Inference with Neural Interpreters](https://mlsanity.com/paper/2110.06399/) (2021)
- [Robustness via Cross-Domain Ensembles](https://mlsanity.com/paper/2103.10919/) (2021)
- [Closer Look at the Transferability of Adversarial Examples: How They Fool Different Models Differently](https://mlsanity.com/paper/2112.14337/) (2021)
- [Causal Balancing for Domain Generalization](https://mlsanity.com/paper/2206.05263/) (2022)
- [Meta Adversarial Training against Universal Patches](https://mlsanity.com/paper/2101.11453/) (2021)
- [Universal Adversarial Perturbations Through the Lens of Deep Steganography: Towards A Fourier Perspective](https://mlsanity.com/paper/2102.06479/) (2021)
- [The worst of both worlds: A comparative analysis of errors in learning from data in psychology and machine learning](https://mlsanity.com/paper/2203.06498/) (2022)
- [Multimodal Scale Consistency and Awareness for Monocular Self-Supervised Depth Estimation](https://mlsanity.com/paper/2103.02451/) (2021)
- [Understanding Bias in Large-Scale Visual Datasets](https://mlsanity.com/paper/2412.01876/) (2024)
- [Toward Learning Human-aligned Cross-domain Robust Models by Countering Misaligned Features](https://mlsanity.com/paper/2111.03740/) (2021)
- [Simplicity Bias Leads to Amplified Performance Disparities](https://mlsanity.com/paper/2212.06641/) (2022)

## Abstract (arXiv)

Deep CNNs are known to exhibit the following peculiarity: on the one hand they generalize extremely well to a test set, while on the other hand they are extremely sensitive to so-called adversarial perturbations. The extreme sensitivity of high performance CNNs to adversarial examples casts serious doubt that these networks are learning high level abstractions in the dataset. We are concerned with the following question: How can a deep CNN that does not learn any high level semantics of the dataset manage to generalize so well? The goal of this article is to measure the tendency of CNNs to learn surface statistical regularities of the dataset. To this end, we use Fourier filtering to construct datasets which share the exact same high level abstractions but exhibit qualitatively different surface statistical regularities. For the SVHN and CIFAR-10 datasets, we present two Fourier filtered variants: a low frequency variant and a randomly filtered variant. Each of the Fourier filtering schemes is tuned to preserve the recognizability of the objects. Our main finding is that CNNs exhibit a tendency to latch onto the Fourier image statistics of the training dataset, sometimes exhibiting up to a 28% generalization gap across the various test sets. Moreover, we observe that significantly increasing the depth of a network has a very marginal impact on closing the aforementioned generalization gap. Thus we provide quantitative evidence supporting the hypothesis that deep CNNs tend to learn surface statistical regularities in the dataset rather than higher-level abstract concepts.