Implements stacked denoising autoencoder in Keras without tied weights.
To read up about the stacked denoising autoencoder, check the following paper:
Vincent, Pascal, Hugo Larochelle, Isabelle Lajoie, Yoshua Bengio, and Pierre-Antoine Manzagol. "Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion." Journal of Machine Learning Research 11, no. Dec (2010): 3371-3408.
Requirements:
- Python 3.4
- keras 1.2
- numpy
- scipy