Agustinus Kristiadi

I think this is the problem on Spacy. Could you try the solution provided here? https://github.com/explosion/spaCy/issues/2015

Hi, it is because in this line there is negative sign. https://github.com/wiseodd/generative-models/blob/c790d2c8add3d600c03ee9a301a969e10ccd7562/GAN/wasserstein_gan/wgan_tensorflow.py#L86 Please do not close this issue as I'd like to keep this open, so other can see in...

Hi, I used ALI paper (https://arxiv.org/pdf/1606.00704.pdf) as reference for this code. You might want to check Algorithm 1 in page 3 there. Cheers

So that we run full ancestral sampling on both cases => independent samples of x. Haven't thought whether this have profound impact or not though. Feel free to experiment.

My point on that implementation is to get the intuition of GibbsNet, and how to quickly implement it, esp. given ALI/BiGAN implementation. Given this intuition feel free to experiment with...

Yeah, it's also what I experienced. It won't generate good result and the loss is blowing up. I suspect we need to have stronger networks + much lower learning rate....

The codes in this repo are actually using the empirical Fisher matrix, so take them with a grain of salt. To compute the true Fisher matrix, you need to do...

Yes, this particular code is for (binary) classification with Bernoulli likelihood. See https://github.com/wiseodd/natural-gradients/blob/7d51f19d315cb393a52a371162aaf2e27d20dbaa/numpy/toy/full_fisher.py#L25-L26

True. The root of this problem is the misspecified loss being used in the code. Natural gradient descent requires us to have a negative-log-likelihood loss, so we should sum the...

Hi, it's basically just pure vanilla Jekyll, plus some plugins (which you can just install easily via gem, e.g. `gem install jekyll-pagination`. Just try do `jekyll serve` and if any...