Hi,

Do you know how to reshape image output by your code in order to get inception score (I guess this code is for coloured images).

here is the inception github code: https://github.com/openai/improved-gan/tree/master/inception_score

Thanks in advance

loss = recon_loss + kl_loss

While you should have loss = recon_loss - kl_loss

To perform gradient descent.

If not you should loss = -recon_loss + kl_loss and perform gradient ascent.

The ELbo to maximize is: ELBO = -DKL + LogLikelihood

Your kl_loss correspond to the -DKL and your binary error to -LogLikelihood. So you have to change your -DKL to be DKL to minimize the ELBO, if not change sign from LogLikelihood and maximize.

Hi,

I think the current implementation for Boundary-seeking GANs is only applicable for continuous variables, and not for discrete which was the main contribution of the paper. Is a discrete implementation in the works?

In the code , did you use Wasserstein to count the distance between discriminator and generator ? I can't see the algorithm of it, could you shed a light ? Thanks

Hi, great work and repo. I have one question about the way the Qloss is computed in infoGAN here ent = tf.reduce_mean(-tf.reduce_sum(tf.log(c + 1e-8) * c, 1)) To me, log(c) is not correct there. I guess it should be log(P(c)), so log(0.1) in this implementation. With that, the Qloss value than reach the 2.3 of H(c) mentioned in the paper. Does it make sense? Thanks

I have used coupled GAN code for my own customize dataset. Discriminator and generator loss not decrease or increase and output noise only. I am unable to understand what is the real problem. Can you please help me ? I have tried to debug it but it generate noise images. please help me! Same generator and discriminator loss.

Hello Wiseodd,

The source code of GAN and its variations provided here are very helpful for those who are interested in applying GAN in bioinformatics.

My question is how to extract encoded features from an AAE (e.g. aae_pytorch.py in the VAE folder). After reading the codes in aae_pytorch.py, I guess that the following codes can be used to extracted the encoded features of training set of MNIST data.

X = Variable(torch.from_numpy(mnist.train.images))
z_fake = Q(X)
np.savetxt('encoded_feature_MNIST.txt', z_fake.data.numpy())

Am I correct?

Thanks,

Xiangchun

fix a mix use of tensor( size 64x1 ) and tensor( size 64 ) which results: UserWarning: Using a target size (torch.Size([64])) that is different to the input size (torch.Size([64, 1])) is deprecated. Please ensure they have the same size.

Hi, If its possible, could you please add our variant of GAN to your collection too? It's a semi-supervised version of infoGAN. Thanks!

https://arxiv.org/abs/1707.04487

Hi, I have got an error while running python infogan_pytorch.py

Extracting ../../MNIST_data/train-images-idx3-ubyte.gz Extracting ../../MNIST_data/train-labels-idx1-ubyte.gz Extracting ../../MNIST_data/t10k-images-idx3-ubyte.gz Extracting ../../MNIST_data/t10k-labels-idx1-ubyte.gz /home/arghyapal/miniconda2/envs/generative-models/lib/python3.5/site-packages/torch/optim/adam.py:74: UserWarning: tensor1 is not broadcastable to self, but they have the same number of elements. Falling back to deprecated pointwise behavior. p.data.addcdiv_(-step_size, exp_avg, denom) Traceback (most recent call last): File "infogan_pytorch.py", line 119, in <module> reset_grad() File "infogan_pytorch.py", line 86, in reset_grad p.grad.data.zero_() AttributeError: 'NoneType' object has no attribute 'data'

Please help me...!!!

Hi, @wiseodd ,

Thanks for your cool implementation of these models!

I'm a newbie to GANs and tensorflow. I've tried vanilla gan by running gan_tensorflow.py.

At the end of training, the content of G becomes all 1-s, ie, all the squares just contain the image of number one, not like the snapshot in the mid-term of the training, all numbers are generated.

Are these results correct? How to explain it? Thanks in advance!

Best regards.

In your blog post you have

Shouldn't that be

I think in your formulation tr(Σ(Χ)) should be tr(Σ^(-1)(Χ)) μ(Χ)^(Τ) μ(Χ) should be μ(Χ)^(Τ) Σ^(-1)(Χ) μ(Χ) -log(det(Σ(Χ))) should be +log(det(Σ(Χ)))

Hey guys, I'm using the conditional-gan code for tensorflow version to generate garment drawings, the training images is RGB png file, labels are vectors like this [1,2,3,4,...,2]，I just changed the feeding training data from mnist to my data, but the generated images are shapeless, I don't know why, someone helps me? thanks!! The attached is the modified code and images generated.

code+generated images.zip

Hello WiseOdd,

I have recognise you vanilla VAE, which seems pretty neat despite of the fact that does not work as I remember a Gassian noise sparse model would work.

I have recently read 1606.05908 where VAE are explained quite good more or less. Threre is the PDF for X described as expectation value over $P(Z)$. Now, when you update your parameters of the Gaussian, normally the variance is a diagonal matrix of $\sigma^2$. So, of cause one could think about a variance for each point in the dimensionality of X, but I am not qute sure that is the proper common idea behind the variance of the Gaussian distribution in generative models.

$P(X) = \frac{1}{D_z} \sum_{\forall z \in Z} P(X|Z, \Theta) P(Z)$ using this annotation, the expectation for $\sigma^2$ based on the data would be something similar to $\sum_{\forall n \in N} \braket{X-\mu(z;\Theta)|X-\mu(z;\Theta)}$, right?

Maybe you can take a look on that code. Because your variance seems to appear as a matrix instead of a value or set of values for the Gaussian noise model.

So, did you thought explicitly about a full covariance matrix or was it just trial and error in this case?

If trying out expectation value(s) for the variance, let me know, what your experience is.

regards, Markus