⭐⭐⭐ Learned Data Augmentation. You could use a learned VAE of data, to perform “learned data augmentation”. One would first train a VAE on input data, then each training point would be transformed by encoding to a latent space, then applying a simple (e.g. Gaussian) perturbation in latent space, then decoding back to observed space. Could we use such an approach to obtain improved generalization? A potential benefit of such data augmentation is that it could include many nonlinear transformations like viewpoint changes and changes in scene lightning. Can we approximate the set of transformations to which the label is invariant? Check out the existing work on this topic if you want a place to get started.
+ +Preliminary paper describing learned data augmentation can be found here , with the implementation avaiable at this repo . Current results improve the generalization of the VAE when tested on the CIFAR-10 dataset.
\ No newline at end of file