Hypothesis of Denoising Autoencoder#
robustness to partial destruction
of the input, i.e., partially destroyed inputs should
yield almost the same representation.such hypothesis is motivated by
a good representation is expected to capture stable structures in the
form of dependencies and regularities characteristic of
the (unknown) distribution of its observed input. For
high dimensional redundant input (such as images) at
least, such structures are likely to depend on evidence
gathered from a combination of many input dimen
sions. They should thus be recoverable from partial
observation only.https://www.cs.toronto.edu/~larocheh/publications/icml-2008-denoising-autoencoders.pdf ↗