Minimal Sufficient Views: A DNN model making predictions with more evidence has higher accuracy

Abstract

Deep neural networks (DNNs) exhibit high performance in image recognition; however, the reasons for their strong generalization abilities remain unclear. A plausible hypothesis is that DNNs achieve robust and accurate predictions by identifying multiple pieces of evidence from images. Thus, to test this hypothesis, this study proposed minimal sufficient views (MSVs). MSVs is defined as a set of minimal regions within an input image that are sufficient to preserve the prediction of DNNs, thus representing the evidence discovered by the DNN. We empirically demonstrated a strong correlation between the number of MSVs (i.e., the number of pieces of evidence) and the generalization performance of the DNN models. Remarkably, this correlation was found to hold within a single DNN as well as between different DNNs, including convolutional and transformer models. This suggested that a DNN model that makes its prediction based on more evidence has a higher generalization performance. We proposed a metric based on MSVs for DNN model selection that did not require label information. Consequently, we empirically showed that the proposed metric was less dependent on the degree of overfitting, rendering it a more reliable indicator of model performance than existing metrics, such as average confidence.

24 pages

References in corpus (14)

• Explaining and Harnessing Adversarial Examples
• scikit-image: Image processing in Python
• Learning Important Features Through Propagating Activation Differences
• Neural Tangent Kernel: Convergence and Generalization in Neural Networks
• Explaining Machine Learning Classifiers through Diverse Counterfactual Explanations
• Understanding deep learning requires rethinking generalization
• Interpretable Explanations of Black Boxes by Meaningful Perturbation
• Captum: A unified and generic model interpretability library for PyTorch
• Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)
• YOLOP: You Only Look Once for Panoptic Driving Perception
• Benchmarking Neural Network Robustness to Common Corruptions and Perturbations
• RISE: Randomized Input Sampling for Explanation of Black-box Models
• Towards Understanding Ensemble, Knowledge Distillation and Self-Distillation in Deep Learning
• What made you do this? Understanding black-box decisions with sufficient input subsets