Learning dynamics of linear denoising autoencoders

Date
2018
Journal Title
Journal ISSN
Volume Title
Publisher
PMLR
Abstract
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.
Description
CITATION: Pretorius, A., Kroon, S. & Kamper, H. 2018. Learning dynamics of linear denoising autoencoders. In Proceedings of the 35 th International Conference on Machine Learning, PMLR 80:4141-4150, 10-15 July 2018, Stockholm, Sweden.
The original publication is available at http://proceedings.mlr.press/
Keywords
Computer science -- Mathematics -- Congresses, Denoising Autoencoders, Computer Science -- Research
Citation
Pretorius, A., Kroon, S. & Kamper, H. 2018. Learning dynamics of linear denoising autoencoders. In Proceedings of the 35 th International Conference on Machine Learning, PMLR 80:4141-4150, 10-15 July 2018, Stockholm, Sweden