What Affects Learned Equivariance in Deep Image Recognition Models?
This work addresses how to enhance equivariance in deep learning models for improved performance in image recognition, though it appears incremental as it builds on existing concepts of equivariance.
The paper tackled the problem of quantifying learned equivariance in neural networks by proposing an improved measure, finding a correlation between learned translation equivariance and validation accuracy on ImageNet, and identifying factors like data augmentation, reduced model capacity, and convolutions that increase learned equivariance.
Equivariance w.r.t. geometric transformations in neural networks improves data efficiency, parameter efficiency and robustness to out-of-domain perspective shifts. When equivariance is not designed into a neural network, the network can still learn equivariant functions from the data. We quantify this learned equivariance, by proposing an improved measure for equivariance. We find evidence for a correlation between learned translation equivariance and validation accuracy on ImageNet. We therefore investigate what can increase the learned equivariance in neural networks, and find that data augmentation, reduced model capacity and inductive bias in the form of convolutions induce higher learned equivariance in neural networks.