Questions tagged [generative-models]

For questions about models designed for generating new data (or generating samples from a probability distribution).

Filter by
Sorted by
Tagged with
0
votes
0answers
7 views

Understanding the ResnetGenerator in the CycleGan Model

In their famous research paper on CycleGans, the authors implement - well, a CycleGan. There are two discriminators and two generators for the CycleGan. Now, they also provide their neural networks ...
0
votes
0answers
10 views

Image parameters for SRGAN

In some implementations of SRGAN I've noticed, that datasets consist of the high-resolution images and the low-resolution images are created later by, e.g. resizing (decreasing the size) hr-images. ...
0
votes
0answers
15 views

How to evaluate pix2pix?

As far as I know, to evaluate synthesized images it is proposed to use: human scoring, "Inception score", where in the second case the quality is rated based on a pre-trained Inception ...
1
vote
0answers
6 views

Training with different datasets for the same better VAE model yields poor results

The VAE model I used here https://github.com/keras-team/keras-io/blob/master/examples/generative/vae.py. It can produce very well results for the minist and fashion minist dataset. But when I use my ...
0
votes
0answers
9 views

Recommendations for viable NN architectures to find learned features? Does this architecture make sense?

My general question is about the best neural network architecture (GAN variant, Autoencoder variant, etc.) that would accomplish the task of outputting the features that a model learned. The idea ...
0
votes
0answers
19 views

DCGAN: why does my generator has less loss then my discriminator?

I have constructed a DCGAN (deep convolutional generative adversarial network) inspired by this github repository. It is written in a more low level Tensorflow code that I tried transforming into ...
0
votes
0answers
12 views

DCGAN: how do we construct the proposed CNN of the original DCGAN paper?

I am reading the original paper on the Deep Convolutional GAN (link: DCGAN paper) and on the fourth page the authors make a proposition on how to model the generator and discriminator as CNNs. However ...
0
votes
0answers
6 views

Audio generating models

I was looking for an implementation of audio\timeseries generation models. I want a model that utilizes time series as well as a frequency map (such as log-Mel or STFT). My goal is to generate a time ...
0
votes
0answers
9 views

Are there good hyperparameter optimization networks for Generative Adversarial Nets?

Finding good hyperparameters, as the learning rate for gradient descent, is crucial for good performance in deep learning. There exist various automatic methods for tasks as segmentation or ...
1
vote
0answers
11 views

What are the motion planning benchmarks?

Suppose I wanted to try and improve on existing motion planning algorithms. What benchmarks should I be trying to beat? Papers with code site has no motion planning benchmarks. I searched online and ...
1
vote
1answer
33 views

Using GANs to generate synthetic tabular data to improve supervised learning

One topic I see some people trying is using GANs to generate synthetic tabular data for supervised learning. Also as a way to oversample the minority class in a binary classification. For me creating ...
0
votes
0answers
10 views

Learning conditional statements from natural language

Natural language text in emails might have conditional statements in them. Are there any technical papers and methods that explore converting unstructured text (eg. emails) into structured conditional ...
2
votes
1answer
15 views

Why a GAN trained on same data and same parameters may produce different results?

I am trying to train a Generative Adversarial Network and ran the training a few times with same dataset and same parameters but it seems tp produce different results. Why this may happen?
2
votes
1answer
36 views

How to build Generative Model when we have more than one variable

I have a data-frame which has looks similar to this: A B C 1 2 2 2 4 3 4 8 5 9 16 7 16 32 11 22 43 14 28 55 17 34 67 20 40 79 23 ...
0
votes
0answers
17 views

Training a Variational Autoencoder (VAE) for Non-Uniform Random Number Generation

I have a complicated 20-dimensional non-uniform distribution and would like generate samples from it. I have considered training a VAE to do so, but my problems are the following: Is my approach ...
1
vote
1answer
58 views

Training a Variational Autoencoder (VAE) for Random Number Generation

I have a complicated 20-dimensional multi-modal distribution and consider training a VAE to learn an approximation of it using 2000 samples. But particularly, with the aim to subsequently generate ...
0
votes
0answers
18 views

Training a VAE for random number generation

I have a high-dimensional multi-modal distribution of random numbers in R^n and consider training a VAE to learn the distribution. What I want to do with it succeedingly, is to sample from the latent ...
2
votes
1answer
77 views

Transformer masking during training or inference?

I'm working through Attention is All you Need, and I have a question about masking in the decoder. It's stated that masking is used to ensure the model doesn't attend to any tokens in the future (not ...
0
votes
1answer
56 views

what is meant by minimizing and maximizing in GANs?

It is a subtle change that involves the generator maximizing the log of the discriminator probabilities for generated images instead of minimizing the log of the inverted discriminator probabilities ...
0
votes
0answers
24 views

Variational Autoencoder (VAE) latent features

I'm new to DL and I'm working on VAE for biomedical images. I need to extract relevant features from ct scan. So I created first an autoencoder and after a VAE. My doubt is that I don't know from ...
0
votes
1answer
15 views

Generate series of values using Keras GAN architecture

I'm trying to generate something like that: Which is a random sample from my real data function (that i'm trying to mimic). ...
0
votes
0answers
13 views

How to build model(s) for simulating conversations between two parties?

I'm working on a problem for chat based buying and selling of products. Due to lack of training data I most likely have to use synthetic data. I was trying to come up with ways to do that, but so far ...
0
votes
0answers
19 views

accessing individual layers from a saved model

I am in the process of writing a model where I have the layers of the model in the "def _init_( )" section of a class I want to save and load the model in the class' variables where i ahve ...
3
votes
1answer
69 views

Help interpreting GAN output, and how to fix it?

After a few tries, I had trained a GAN to produce semi-sensible output. In this model, it almost instantly found a solution and got stuck there. The loss for both the discriminator and generator were ...
1
vote
0answers
14 views

Checkerboard artefacts vs distinct objects in GANs

I found a very good solution for getting rid of checkerboard artefacts in GANs: https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/issues/190 Instead of using Transposed Convolution, use bilinear ...
1
vote
0answers
19 views

Magenta MusicVAE/GrooVAE conditioning

I want to try different methods of conditioning the decoding process of the Variational Autoencoder Models of the Google Magenta project for my own research project. As far as I can tell, MusicVAE has ...
1
vote
0answers
44 views

Fine tuning Conditional GANs for low data scenarios

I was wondering what the process was for fine tuning a conditional GAN. For example, say I wanted to generate pictures of an object X given a certain condition such as a sentence describing it, which ...
0
votes
0answers
8 views

what are Latent model and Latent Varibles?

what are latent models and why they use Expectation maximization instead of gradient descent to optmize the parameters ? How to interpret and understand whether a hidden varible is present in model or ...
1
vote
0answers
111 views

MNIST GAN generator loss increasing

I'm trying to train a simple vanilla GAN on MNIST with Tensorflow. I used this github page as a reference and in my process to try and get my GAN to work I've made my code more and more similar to ...
0
votes
0answers
14 views

During modeling binarized handwritten digit images using a BMM and EM algorithm, why must, during initialization, the value of means sum to 1?

This is my first question on StackExchange. So please pardon any mistakes that I might've made. I'm reading about Multivariate Bernoulli Mixture Models and the EM algorithm from Pattern Recognition ...
0
votes
1answer
225 views

How to make custom callback in keras to generate sample image in VAE training?

I'm training a simple VAE model on 64*64 images and I would like to see the images generated after every epoch or every couple batches to see the progress. when I train the model I wait until the ...
0
votes
1answer
1k views

Is it possible to train stylegan2 with a custom dataset using a graphics card that only has 6GB of VRAM (GeForce GTX 1660)?

I'm attempting to train stylegan2 using a custom dataset, but no matter what settings I use I see the same error: ...
0
votes
0answers
8 views

generative models visualizations

Can someone recommend me a paper or blog where they show nice visualizations of reconstructed data in a generative model, which is not trained with pictures. Visualization of reconstructed images is ...
0
votes
0answers
15 views

WGAN training tradeoffs

Currently training a WGAN with weight clipping and having reread the architecture and pitfalls mentioned in code, I am running it with 2 layers in critic and generator, no batch norm in generator and ...
0
votes
0answers
21 views

Order of importance of hyperparameters

I'm trying to make a generative model and for a given input shape there is a fixed corresponding output shape. Because of the way convolutional and transposed convolutional layers determine output ...
0
votes
0answers
15 views

Using DCGAN on a (very small) dataset of art

I am developing a DCGAN using the this tutorial in PyCharm. As my usage of this tutorial suggests, I am quite new to DCGANs as I've previously only had a few experiences with machine learning ...
0
votes
0answers
37 views

Keras: Implementing paired Data Augmentation into Pix2Pix cGAN confusion

but learning GANs. Following a Jason Lee Brown tutorial (the code is his)... I am trying to add in Keras' ImageDataGenerator to the pix2pix algorithm. I can ...
1
vote
0answers
16 views

Limits of generative models

If I was to train a pixel rnn, by concatenating images with noise vectors, I should be able to get new images from that distribution by starting on the prediction with a new noise vector. Would it be ...
1
vote
0answers
123 views

Why -logD trick in GANs training work?

When training the generator, we want to minimize log(1 - D(G(z))) (this has a form of log (1-x)). Using graph calculator we plot and see that the slope of this function is very small at first, thus ...
1
vote
0answers
38 views

Is the hyperbolic tangent function a solution to the weight clipping problem of WGAN?

Instead of clipping to the range $[-c,c]$ in WGAN (Wasserstein generative adversarial network), why not smoothly map into the range $[-c,c]$ by using $c\times \mathrm{tanh}(w)$? This would guarantee ...
1
vote
0answers
52 views

How can I know my GAN does converge?

I am training my GAN, and by looking at the loss curves of the generator and discriminator, I think I am going on the right way because from this blog, my curves look reasonable: A stable GAN will ...
2
votes
1answer
2k views

Inception Score (IS) and Fréchet Inception Distance (FID), which one is better for GAN evaluation?

IS uses two criteria in measuring the performance of GAN: The quality of the generated images, and their diversity based on the entropy of the distribution of synthetic data. On the other hand, FID ...
6
votes
3answers
807 views

Is it possible to use a generative model to “share” private data?

Let's say we have some data set, with lots of instances $X$ and a target $y$. If it is of some importance, you may assume that it is a "real life" data set : medium sized, with important correlations, ...
1
vote
0answers
39 views

Reinforcement learning in bidirectional RNN

I have been self-learning deep generative neural network for a while. I am okay with the basics but I really need some guidance and jump start. I have recently came across this paper “Bidirectional ...
1
vote
2answers
321 views
1
vote
1answer
97 views

GAN: Discriminator converges, generator learns almost nothing

In my GAN, the discriminator loss goes down steadily, while the generator loss oscillates / does not converge. I suspect this is due to the vanishing gradient problem. Theory: as the discriminator ...
1
vote
1answer
358 views

CycleGAN: Generator losses don't decrease, discriminators get perfect

So I´m training a CycleGAN for image-to-image transfer. The problem is: while the discriminator losses decrease, and are very small now, the generator losses don't decrease at all. The generator ...
2
votes
1answer
50 views

Create an RNN on text sources with different lengths

I want to create an RNN to generate a new text based on many examples of existing texts of a certain format in the training data. The type of texts in the training data consists of 3 segments, like so:...
2
votes
0answers
33 views

Generative Model for learning Periodic solutions 3-body/N-body problems

I am tasked with finding research where a GAN or any other generative model is used to generate new shapes of 3-bodies moving under the influence of each others' gravitational pull, in a periodic ...
3
votes
1answer
30 views

How to supress previous results in a generative network?

Think of a generative network generating animal names. It's been trained on a set of animal names so this shouldn't be too hard. But say I want to generate 10 animal names. First I run the network ...