Analyzing noise in autoencoders and deep networks
This work provides a unified noise injection framework for autoencoders, which is incremental as it builds on existing methods to enhance representation learning in unsupervised settings.
The authors tackled the problem of improving autoencoder performance by extending denoising autoencoders to inject noise at multiple stages, showing that this unified framework outperforms denoising autoencoders in denoising tasks and is competitive on benchmarks like MNIST and CIFAR-10.
Autoencoders have emerged as a useful framework for unsupervised learning of internal representations, and a wide variety of apparently conceptually disparate regularization techniques have been proposed to generate useful features. Here we extend existing denoising autoencoders to additionally inject noise before the nonlinearity, and at the hidden unit activations. We show that a wide variety of previous methods, including denoising, contractive, and sparse autoencoders, as well as dropout can be interpreted using this framework. This noise injection framework reaps practical benefits by providing a unified strategy to develop new internal representations by designing the nature of the injected noise. We show that noisy autoencoders outperform denoising autoencoders at the very task of denoising, and are competitive with other single-layer techniques on MNIST, and CIFAR-10. We also show that types of noise other than dropout improve performance in a deep network through sparsifying, decorrelating, and spreading information across representations.