Learning Dynamics of Linear Denoising Autoencoders
This work provides theoretical insights into the regularization effects of noise in DAEs, which is an incremental advancement for researchers in unsupervised representation learning.
The authors tackled the problem of understanding how input noise influences learning in denoising autoencoders (DAEs) by developing a theoretical framework for linear DAEs, deriving analytic expressions that describe their learning dynamics and verifying these with simulations and experiments on datasets like MNIST and CIFAR-10.
Denoising autoencoders (DAEs) have proven useful for unsupervised representation learning, but a thorough theoretical understanding is still lacking of how the input noise influences learning. Here we develop theory for how noise influences learning in DAEs. By focusing on linear DAEs, we are able to derive analytic expressions that exactly describe their learning dynamics. We verify our theoretical predictions with simulations as well as experiments on MNIST and CIFAR-10. The theory illustrates how, when tuned correctly, noise allows DAEs to ignore low variance directions in the inputs while learning to reconstruct them. Furthermore, in a comparison of the learning dynamics of DAEs to standard regularised autoencoders, we show that noise has a similar regularisation effect to weight decay, but with faster training dynamics. We also show that our theoretical predictions approximate learning dynamics on real-world data and qualitatively match observed dynamics in nonlinear DAEs.