Re:Autoencoding Blade Runner: reconstructing films with arti...
by bcheung on Tuesday May 24, 2016 @09:31PM
(#11767048)
Very interesting. This makes me wonder if a similar technique can be used for compression?
[ Parent
| on HN ]
Re:Autoencoding Blade Runner: reconstructing films with arti...
by aab0 on Tuesday May 24, 2016 @11:38PM
(#11767507)
Absolutely. The point of an autoencoder is dimensionality reduction: boil a big set of data down to a few hundred or thousand numbers in a vector which summarizes it. You could treat it either as lossy compression and store just the encoding, or you can treat it as a hybrid format in which the autoencoder lossy encoding is then corrected to lossless by additional bits in the stream.
In practice, even the hyper-efficient compression algorithms used in something like zpaq tend to use only very small shallow predictive neural networks because no one wants to wait days for their data to be compressed or ship around big neural nets as part of their archives, so it's more of an information-theoretic curiosity. Few enough people will even use 'xz'.
Re:Autoencoding Blade Runner: reconstructing films with arti...
by vintermann on Wednesday May 25, 2016 @01:58AM
(#11767922)
You could, but I don't think this would be competitive on compression ratio at all, even allowing for an order of magnitude more time.
Re:Autoencoding Blade Runner: reconstructing films with arti...
by sp332 on Wednesday May 25, 2016 @10:13AM
(#11769877)
The technique is very effective, after all a variant of paq holds a compression record. http://prize.hutter1.net/ You can try paq yourself in PeaZip. http://filehippo.com/download_peazip_64/
Re:Autoencoding Blade Runner: reconstructing films with arti...
by argonaut on Wednesday May 25, 2016 @05:32PM
(#11773299)
PAQ does not use autoencoders, that's the difference. Neural network models for compression are extremely slow and can use a crapton of memory (one top benchmark result uses 32GB of RAM): https://www.quora.com/What-is-the-potential-of-neural-networ...
Re:Autoencoding Blade Runner: reconstructing films with arti...
by vintermann on Thursday May 26, 2016 @02:50AM
(#11776099)
Last I checked, PAQ only uses a shallow (two layer) neural network as a last step to weight the predictions from the multiple handmade next-bit prediction models it contains.
Re:Autoencoding Blade Runner: reconstructing films with arti...
by sp332 on Thursday May 26, 2016 @11:07AM
(#11778475)
So, you think the additional accuracy of the larger neural network isn't enough to overcome the storage size of the network itself?
Re:Autoencoding Blade Runner: reconstructing films with arti...
by csn on Wednesday May 25, 2016 @01:03AM
(#11767750)
I was intrigued by this a while back. I think of training a NN as generating a function, an equation, from a training set which, given a specific input, outputs a prediction. If you can come up with an equation-input pair that, when executed, a) accurately enough approximates some data, and b) requires less space than the original file, you have achieved (most likely lossy) compression.
Re:Autoencoding Blade Runner: reconstructing films with arti...
by astrange on Wednesday May 25, 2016 @02:29AM
(#11768000)
This is too powerful to use for product work, I think.
Professionals like to use something they can understand, and when making a BD or streaming source they know what a compression artifact looks like and which frames' bitrates to tweak to hide it. They pretty much sit there all day and just do that.
Maybe live streaming?
Re:Autoencoding Blade Runner: reconstructing films with arti...
by adriancooney on Tuesday May 24, 2016 @09:52PM
(#11767132)
Middle-out compression.