Autoencoders were first introduced as a method for unsupervised pre-training in Modular Learning in Neural Networks (D. Ballard, AAAI proceedings, 1987). They were then used for dimensionality reduction, such as in Auto-Association by Multilayer Perceptrons and Singular Value Decomposition (H. Bourlard and Y. Kamp, biological cybernetics, 1988; 59:291-294) and non-linear feature learning, for example, Autoencoders, Minimum Description Length, and Helmholtz Free Energy (G. Hinton and R. Zemel, Advances In Neural Information Processing Systems, 1994).
Autoencoders have evolved over time and there have been several variants proposed in the past decade. In 2008, P. Vincent et al. introduced denoising autoencoders (DAEs) in Extracting and Composing Robust Features with Denoising Autoencoders (proceedings of the 25th International Conference...