recentpopularlog in

deeplearning

« earlier   
Are Energy-Based GANs any more energy-based than normal GANs?
EDIT: The post had a mistake, pretty crucial one, kindly pointed out by George Tucker, so the has changed from what it was in the original. I have falsely claimed that a variant of the general GAN algorithm is pathological, this turns out not to be the case. Sorry for the mistake.
Archive  deeplearning  machinelearning 
10 hours ago by ronert
How powerful are Graph Convolutions? (review of Kipf & Welling, 2016)
This post is about a paper that has just come out recently on practical generalizations of convolutional layers to graphs: Along the way I found this earlier, related paper: This post is mainly a review of (Kipf and Welling, 2016).
Archive  deeplearning  machinelearning 
10 hours ago by ronert
InfoGAN: using the variational bound on mutual information (twice)
Many people have recommended me the infoGAN paper, but I hadn't taken the time to read it until recently. It is actually quite cool: The InfoGAN idea is pretty simple. The paper presents an extension to the GAN objective.
Archive  deeplearning  machinelearning 
10 hours ago by ronert
Understanding Minibatch Discrimination in GANs
There was one idea in there which got me thinking, and this is what I wanted to write about here: minibatch discrimination. Here is the ipython/jupyter notebook I used to draw the plots and test some of the things in this post in practice.
Archive  deeplearning 
10 hours ago by ronert
Dilated Convolutions and Kronecker Factored Convolutions
These are my notes on an ICLR paper from this year: Whilst I wrote this note I also became aware of this paper: I think the two are related, but coming at the same thing from two different directions.
Archive  deeplearning  machinelearning 
10 hours ago by ronert
Adversarial Preference Loss
In my earlier post I talked about how the GAN training procedure can be modified slightly so it minimises $KL[Q\|P]$ divergence. I have also talked about how GANs can be understood as
Archive  deeplearning 
10 hours ago by ronert
An Alternative Update Rule for Generative Adversarial Networks
It is mentioned in the original GAN paper (Goodfellow et al, 2014) that the algorithm can be interpreted as minimising Jensen-Shannon divergence under some ideal conditions.
Archive  deeplearning  machinelearning 
10 hours ago by ronert

Copy this bookmark:





to read