Create Gaussian Noise layer

[Reddine]

Reference : Vincent, Pascal, et al. "Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion." The Journal of Machine Learning Research 11 (2010): 3371-3408.

[fchollet]

Cool! I'm wondering if this belongs among the core layers, though. I remember that at some point we discussed the possibility of introducing a "noise" module. If there was such a module, what other layers would be in it? (I like having Dropout in core, though).

[Reddine]

may be as mentioned in the Reference article: Masking Noise (same principle of Dropout) and Salt-and-pepper Noise.

[pranv]

Neat!
I think that Noise can have its own module. The paper - Deep Learning with Noise has many versions of noise - not just Gaussian. Having the entire family might be useful.

[fchollet]

I put it in a noise module.