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
15 views

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

Instead of clipping to the range [-c,c] why not smoothly map into the range [-c,c] by using c*tanh(w) ? This would guarantee the Lipschizt constant is no greater than c. The problem I am talking ...
1
vote
0answers
15 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 ...
1
vote
1answer
19 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 ...
1
vote
0answers
17 views

Instance segmentation and scene reconstruction

I've been recently interested in various segmentation tasks especially instance segmentation. Having been experimenting with various different datasets I stumbled upon using the method for indoor ...
6
votes
3answers
782 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
24 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 ...
0
votes
0answers
7 views

Virtual Generation of Synthetic Ancient and Dirty English Documents

I have a collection of dirty background image, below is the sample: I have also a collection of an actual image of a dirty document with text on it, just like below: My problem with my actual image ...
1
vote
0answers
24 views

Keras Encoder - Decoder Model

Im training a model of 400 samples . The dataset contains 400 images of faces as input (X) and also 400 faces with glasses as output (Y) . im training the model by code below : ...
1
vote
1answer
28 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
0answers
14 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 ...
1
vote
0answers
18 views

Implementing an RNN on multiple text sources

I want to implement 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, ...
2
votes
0answers
28 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
27 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 ...
2
votes
1answer
26 views

Why joint probability in generative models?

I have been reading about generative models for past few days and there's one question that's bugging me. Many sources(Wiki, Google dev article) mention that generative models try to model the joint ...
5
votes
4answers
61 views

Could GANs be used to augment data?

I want to use GAN for data augmentation but I am confuse what are the pros. and cons. of data augmentation using GAN or why we use data augmentation using GAN compared to other data augmentation ...
0
votes
1answer
57 views

Are there some research papers about text-to-set generation?

I have googled but find no results. Text-to-(word)set generation or sequence-to-(token)set generation. For example, input a text and then output the tags for this text: ...
2
votes
0answers
18 views

Are Denoising Variational Autoencoders deterministic?

I have a pretty good understanding of regular autoencoders and, to a certain extent, of variational autoencoders, where the latent representation is forced to follow specific probabilistic ...
1
vote
1answer
45 views

Discriminator of a Conditional GAN with continuous labels

OK, let's say we have well-labeled images with non-discrete labels such as brightness or size or something and we want to generate images based on it. If it were done with a discrete label it could ...
1
vote
1answer
45 views

How does joint probability generate new data in generative model

Learning about different Machine Learning concepts, I came across Generative and Discriminative model. To infer from what I have studied, generative model is based on P(x,y)(Joint probability ...
2
votes
1answer
52 views

Time Series Generation - Multi Dimensional Time Series Data

Disclaimer: Mathematicians please don't be mad at me for the use of some of the terminologies in this post. I am an Engineer. :-) Background: So I am currently working on a problem where I have to ...
0
votes
0answers
5 views

Help regarding generative modeling

I have data in form of n dimensional points. I have to train generative model where i pick a class through multinoulli distribution. Then each dimension of actual point is generated by some ...
0
votes
0answers
19 views

Learning discrete probability distribution that is parametrized by a set of real-valued parameters

Assume I have a discrete probability distribution defined over binary variables. This probability distribution is parametrized by a set of real-valued parameters, which all are contained in a segment, ...
0
votes
0answers
18 views

How to solve a supervised learning problem with a generative model?

Is there a framework to do supervised task in generative model fashion? i.e. modelling p(x,y) rather than p(y|x) as in discriminative models. When I look at generative models, they all revolve ...
2
votes
0answers
35 views

Why are discriminative models denoted as P(Y|X?

In general, we can speak we know deterministic and probabilistic models. Discriminative ones are deterministic, while generative one is probabilistic. But I have been reading about the difference ...
0
votes
1answer
36 views

What is a latent space vector?

I do not understand this about GANs. Apparently the Generator is supposed to receive a latent space vector as its input. Yet I couldn't find an example of how I can implement it in Pytorch. This is a ...
1
vote
0answers
13 views

Image Quality Assessment Algorithms

I am working on super resolution projects using Generative Adversarial Networks. I need to validate the generated image whether they are close to original one. I already worked with SSIM and PSNR. is ...
0
votes
1answer
19 views

Why does the generator produce output in a different scale than the training sample?

I am currently trying to train a GAN (based on proGAN) to produce images in a Vaporwave-style which is quite distinct. The results so far have been underwhelming, which I suspect might be due to a ...
2
votes
0answers
18 views

Rescaling simulations by machine learning

I am working in the field of astrophysics. A problem that I consistently find in my projects is that I am always limited by the size of the sky simulations available for a specific observable that I ...
1
vote
1answer
11 views

Is it faster and better to train a GAN on just one digit as opposed to the whole mnist dataset?

When going through an introductory GAN tutorial to generate mnist like handwritten digits I wondered whether the systematic variance in the training data due to the different digits makes the model ...
0
votes
0answers
12 views

How Cycle GAN translates between very different objects?

I'm trying to understand how the popular CycleGAN responds if the objects to be translated between are very different (horse and map or house and apples). All of the examples appear to be translating ...
0
votes
1answer
183 views

WGAN-GP slow critic training time

I am implementing WGAN-GP using Tensorflow 2.0, but each training iteration of the critic is very slow (about 4 secs on my CPU, and somehow 9 secs on Colab GPU). Is WGAN-GP usually this slow or ...
0
votes
0answers
7 views

How to grasp the full entropy of the distribution we want to model in GAN

In pix2pix GAN paper( https://arxiv.org/abs/1611.07004), authors found that the noise vector and the dropout are not efficient in grasping the full entropy of the data distribution we want to model. ...
0
votes
0answers
15 views

InfoGAN Learning Latent Categorical Code

While reading the InfoGAN paper and implement it taking help from a previous implementation, I'm having some difficulty understanding how it learns the discrete categorical code. The implementation ...
0
votes
1answer
305 views

Fully Convolutional Variational Autoencoder

I would like to make a neural network which uses black and white images as input and outputs a colored version of it. The important thing in that process is that the size of the images must stay the ...
0
votes
0answers
18 views

What are the similarities and differences between Imputation, Generative Models and Bootstrapping?

What are the similarities and differences between the following methods: Data Imputation, Generative Models, Bootstrapping.
0
votes
0answers
18 views

Why logarithm in the loss function of GAN

In the official paper of GAN by Ian Goodfellow, while maximizing the probabilities of generator and discriminator, why we take logarithm of the output of Discriminator.
1
vote
1answer
182 views

Why Generator error increases after some epoch

I'm training Realtivistic GAN on dog images. I don't understand why my generator loss increased after 100 epoch. This is the generator loss graph.
0
votes
2answers
128 views

Why does discriminiator accuracy falls to 0%, and is there a fix around this?

I am training a Vanilla-GAN(or original GAN 2016) on a pokemon dataset https://www.kaggle.com/kvpratama/pokemon-images-dataset, for few epochs the discriminator has 100% accuracy over the real ...
1
vote
0answers
17 views

How can one be assured that generative models are not memorizing dataset, and that they will generate an unique image outside of dataset?

If all GAN can do is capture the probability distribution of the dataset, then shouldn't they be similar to handing out images from the dataset? How can we verify that the images that they generate ...
0
votes
0answers
8 views

Post Training Sampling from a unit Gaussian?

During the training of a VAE, we sample Z from N(mu, sigma) ( = Q(Z|X) ) and then feed it to the decoder for reconstruction. Why do we, then, feed Z ~ N(0,1) to the decoder after the training is ...
0
votes
1answer
26 views

What does this symbol means, what operator is it?

I am confused about the $E_{x\sim P_{data}(x)}$, what does $E$ means here. I cannot find an appropriate answer on the internet, and hence I am trying data science stack exchange. Please help.
1
vote
1answer
213 views

Are mainstream pre-trained models useful as discriminators?

In the context of GANs I see many papers designing new discriminator networks. I'm curious about the usefulness of designing discriminators as modified versions of mainstream models like Inception, ...
0
votes
0answers
21 views

Problem regarding designing the generator function

I am implementing the paper Perceptual GAN for small object detection. The design is described by the picture given below. I have used Transfer Learning concept and used a pretrained model inside the ...
1
vote
2answers
200 views

How discriminator knows if the image is real or fake at the initial phase?

If the discriminator and generator in GAN learn together, How the discriminator knows whether the image is real or fake in the initial phase of training? Does the discriminator need to get trained ...
1
vote
1answer
375 views

Comparsion between DCGAN and WGAN

What is the main architectural difference between DCGAN and WGAN? For which problems each models can be more useful than the other one?
1
vote
1answer
1k views

what is the main difference between GAN and autoencoder?

what is the main difference between GAN and other older generative models? what were the characteristics of GAN that made it more successful than other generative models?
0
votes
0answers
18 views

DC GAN to W GAN conversion

I have implemented DC GAN network and I know that changing the loss function , we get a W-GAN network, but I wonder how to code the wasserstein loss function and integrate it with my code below: Here ...
0
votes
0answers
14 views

issues related to implementation of the paper 'Perceptual GAN for small object detection'

I am trying to implement the perceptual gan network using keras as backend. I am new in the field of GAN, can anyone provide me the model for perceptual GAN in keras backend
0
votes
1answer
47 views

Issues related to the code for ROI pooling from the feature map

I am trying to do ROI pooling on the feature map obtained from the VGG layers but I don't know how to code this layers. Can anybody help me out? Here is my VGG layers: ...