Diagnosing Convolutional Neural Networks using their Spectral Response

TL;DR

This paper investigates the spectral response of CNNs and demonstrates that their gain can serve as a diagnostic tool for training issues, potentially replacing validation loss in some scenarios.

Contribution

It introduces the spectral gain as a new diagnostic metric for CNN training, showing its correlation with model performance and overfitting.

Findings

Best models have highest spectral gain sensitivity.

Gain rises during learning and saturates at convergence.

Gain fluctuations indicate overfitting and learning problems.

Abstract

Convolutional Neural Networks (CNNs) are a class of artificial neural networks whose computational blocks use convolution, together with other linear and non-linear operations, to perform classification or regression. This paper explores the spectral response of CNNs and its potential use in diagnosing problems with their training. We measure the gain of CNNs trained for image classification on ImageNet and observe that the best models are also the most sensitive to perturbations of their input. Further, we perform experiments on MNIST and CIFAR-10 to find that the gain rises as the network learns and then saturates as the network converges. Moreover, we find that strong gain fluctuations can point to overfitting and learning problems caused by a poor choice of learning rate. We argue that the gain of CNNs can act as a diagnostic tool and potential replacement for the validation loss when hold-out validation data are not available.