Trivial or impossible -- dichotomous data difficulty masks model differences (on ImageNet and beyond)

Abstract

"The power of a generalization system follows directly from its biases" (Mitchell 1980). Today, CNNs are incredibly powerful generalisation systems -- but to what degree have we understood how their inductive bias influences model decisions? We here attempt to disentangle the various aspects that determine how a model decides. In particular, we ask: what makes one model decide differently from another? In a meticulously controlled setting, we find that (1.) irrespective of the network architecture or objective (e.g. self-supervised, semi-supervised, vision transformers, recurrent models) all models end up with a similar decision boundary. (2.) To understand these findings, we analysed model decisions on the ImageNet validation set from epoch to epoch and image by image. We find that the ImageNet validation set, among others, suffers from dichotomous data difficulty (DDD): For the range of investigated models and their accuracies, it is dominated by 46.0% "trivial" and 11.5% "impossible" images (beyond label errors). Only 42.5% of the images could possibly be responsible for the differences between two models' decision boundaries. (3.) Only removing the "impossible" and "trivial" images allows us to see pronounced differences between models. (4.) Humans are highly accurate at predicting which images are "trivial" and "impossible" for CNNs (81.4%). This implies that in future comparisons of brains, machines and behaviour, much may be gained from investigating the decisive role of images and the distribution of their difficulties.

Published as a conference paper at ICLR 2022

References in corpus (17)

• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• Learning Transferable Visual Models From Natural Language Supervision
• The Space of Transferable Adversarial Examples
• Similarity of Neural Network Representations Revisited
• Underspecification Presents Challenges for Credibility in Modern Machine Learning
• Do ImageNet Classifiers Generalize to ImageNet?
• Billion-scale semi-supervised learning for image classification
• Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs
• Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth
• Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks
• Are we done with ImageNet?
• From ImageNet to Image Classification: Contextualizing Progress on Benchmarks
• What shapes feature representations? Exploring datasets, architectures, and training
• Accelerating Deep Learning by Focusing on the Biggest Losers
• On the surprising similarities between supervised and self-supervised models
• Model Similarity Mitigates Test Set Overuse
• Paradox in Deep Neural Networks: Similar yet Different while Different yet Similar

Cited by in corpus (1)

• Partial success in closing the gap between human and machine vision