Re:Inceptionism: Going Deeper into Neural Networks
by gradys on Thursday June 18, 2015 @12:44PM
(#9739856)
So in concrete terms, does this mean that we show the network an image, choose one layer's output vector, and then back-propagate gradients to the image such that the direction of that vector stays the same, but the magnitude increases?
Re:Inceptionism: Going Deeper into Neural Networks
by murbard2 on Thursday June 18, 2015 @12:46PM
(#9739869)
That is my understanding of the blog post, yes.
That plus a prior on the input pixels to keep it image-like.
Re:Inceptionism: Going Deeper into Neural Networks
by rndn on Thursday June 18, 2015 @01:46PM
(#9740392)
They've written though that they have chosen a particular layer in the network, which reads like "independent of the output layer". Features in such a layer correlate with certain classes, but I don't think they have dealt with classes at all. If that's the case, then the question is how they've amplified the detected features.
Re:Inceptionism: Going Deeper into Neural Networks
by murbard2 on Thursday June 18, 2015 @01:52PM
(#9740432)
Yes they play with various layers. Layers closer to the input act more like edge enhancers, while higher layers emphasize whole objects ("animal" enhancers). You get increasingly less syntactical and increasingly more semantic as you go deeper in the network.
They say: oh you think that cloud is a tiny bit dog-like? Ok, well then find me a small modification to the image that would make it a little more dog like, then a little more, and so on.
Think of it as semantic contrast enhancement