Questions tagged [vae]
The vae tag has no usage guidance.
31
questions
4
votes
0answers
48 views
1D CNN Variational Autoencoder Conv1D Size
I am trying to create a 1D variational autoencoder to take in a 931x1 vector as input, but I have been having trouble with two things:
Getting the output size of 931, since maxpooling and upsampling ...
0
votes
0answers
13 views
Variational autoencoders - encoder-decoder neural nets relationship to maximizing evidence
I am new to VAE and I do not know why if the encoder and decoder are learned by a neural networks then how is maximizing the ELBO (evidence lower bound) or maximizing evidence in general relevant to ...
0
votes
0answers
21 views
VAE KL-divergence with non-standard mean
I know I can make a VAE do generation with a mean of 0 and std-dev of 1.
I tested it with the following loss function:
...
0
votes
0answers
8 views
Reversing data through a tensorflow feature_column.embedding_column
I am building an variational autoencoder in Tensorflow, and one of my columns has object data. The data is too sparse (on the order of 2^16 possible values) to use one-hot encoding, it's not ordinal ...
0
votes
0answers
9 views
VAE: first hidden layer of encoder having more neurons than the input layer
Many blog posts explain the architecture of variational autoencoders (VAEs) with symmetrical encoder-decoder architecture where the number of neurons in the encoder's hidden layers decreases, starting ...
0
votes
0answers
10 views
Single real number evaluation metric for VAE for a regression problem
I've set up a VAE for a regression problem.
...
0
votes
0answers
10 views
Weird cutdown in VAE mean graph
I'm experimenting with VAE with Tensorflow, and just following tutorial setups to have an approximate comparison. Well, they aren't similar at all.
Tutorial mean graph:
My mean graph:
As you can see ...
2
votes
1answer
27 views
What are the Most Dissimilar MNIST Digits?
Using whatever definition of dissimilarity over sets that you'd like, what are the most dissimilar two digits in MNIST? I was thinking that a reasonable approach to answering the question would be to ...
1
vote
0answers
49 views
pytorch code for VAE for MINST forces mu and logvar to zero
I am new to pytorch and trying to implement a VAE for MNIST data. When I try to train my model, it appears that the model forces mu and logvar to zero (or something very close to zero) independent of ...
1
vote
0answers
32 views
Is vanishing KL a problem in vision based VAEs as well?
I came across some work on the problem of a vanishing KL contrbution in Variational Auto Encoders Cyclical Annealing Schedule: A Simple Approach to Mitigating KL Vanishing. This work particularly is ...
3
votes
1answer
70 views
Why KL Divergence instead of Cross-entropy in VAE
I understand how KL divergence provides us with a measure of how one probability distribution is different from a second, reference probability distribution. But why are they particularly used (...
0
votes
0answers
31 views
Varitional Autoencoder not accepting batch size or validation data
The input to the VAE will be a customer vector where the index of the vector represents a product id, position i in vector x is set to one iff product id i has been purchased by the customer. For ...
0
votes
0answers
17 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 ...
0
votes
0answers
111 views
MSE loss in VAE reduces only KL divergence
I am trying to reproduce the two stage VAE from this repository in TF2 with Keras to learn MNIST digits. Unfortunately I am experiencing behavior that I can not explain to myself:
As far as I ...
0
votes
1answer
47 views
What makes the posterior intractable?
In the setting of Variational AutoEncoders, i.e. when we want to find the posterior distribution
over the data generating, latent variable z, given some ...
0
votes
0answers
157 views
Autoencoder works but VAE does not - KL divergence loss drops to zero
I have an image encoding problem, to which I have successfully applied an autoencoder. But when I try to convert this into a VAE framework by introducing KL divergence loss and sampling from a unit ...
0
votes
0answers
73 views
Can the input and output be different in a variational autoencoder?
While the autoencoder and the variational autoencoder typically decode back to the same input space that was passed in, I am working on a more generic encoder-decoder structure, where the output is a ...
3
votes
1answer
170 views
VAE generates bad images. due to unbalanced loss functions?
I'm training a variational autoencoder on CelebA dataset using TensorFlow.keras
The problem I'm facing is that the generated images are not diverse enough and look kinda bad.
(new) Example:
What I ...
0
votes
1answer
181 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
0answers
4 views
What is achieved by converting the latent space to normal distribution in VAE?
Instead of forwarding the latent values to the decoder directly, VAEs use them to calculate a mean and a standard deviation. The input to the decoder is then sampled from the corresponding normal ...
9
votes
1answer
383 views
Train a GAN on “before and after” images of dental surgeries [closed]
I want a GAN to train on "before and after" images of dental surgeries; so that it can generate "after" pictures for fresh patients.
Input images are like these: https://img.webmd.com/dtmcms/live/...
2
votes
2answers
41 views
Can VAEs be used to generate multivariate data?
Most of the tutorials online seem to use VAEs to generate images and use CNNs to generate data. I am working on a game with multivariate data consisting of character position and the character ...
1
vote
0answers
136 views
Variational Autoencoder: Negative log likelihood not optimized
I am using the auto encoding variational Bayes algorithm for one unsupervised object detection task. In the loss function, the reconstruction loss is calculated as the log likelihood of the original ...
1
vote
0answers
63 views
How to estimate total correlation KL[q(z)||Πjq(zj)] of VAE after training (useful for latents disentanglement evaluation)
FactorVAE and β-TCVAE both use total correlation (TC) batch estimation for their objectives. Where TC is:
$$
KL\bigl( q(z)||\prod\nolimits_{j} q(z_{j})\bigr)
$$
both estimates are applied to $q(z|x)$...
0
votes
2answers
65 views
What is the meaning of “probability distribution of p(x)” of something uncountable?
I'm studying VAE and new to both of the neural network and the statistic.
After some researches, I could understand the rough concept of VAE.
But what makes me confused is, the meaning of probability ...
2
votes
1answer
474 views
Intractability in Variational Autoencoders
I'm having difficulty understanding when integrals are intractable in variational inference problems.
In a variational autoencoder with observation $x$ and latent variable $z$ we want to maximize data ...
5
votes
1answer
2k views
Why maximize ELBO in the variational autoencoder?
For a variational autoencoder, we have that:
$$\mathcal{L}(x,\theta,\phi) := \mathbb{E}_{z \sim q_\phi(z|x)}[\log p_{\theta}(x|z)] -KL[q_{\phi}(z|x) ||p(z)] $$
This is called the variational lower ...
2
votes
1answer
2k views
ValueError: Cannot convert a partially known TensorShape to a Tensor: (?, 256)
I'm working on a sequence to sequence approach using LSTM and a VAE with an attention mechanism.
...
2
votes
1answer
2k views
What do we visualize in showing a VAE latent space?
I am trying to wrap my head around VAE's and have trouble understanding what is being visualized when people make scatter plots of the latent space. I think I understand the bottleneck concept; we go ...
0
votes
0answers
1k views
InvalidArgumentError: incompatible shapes: [32,153] vs [32,5] , when using VAE
I'm working on a sequence to sequence model using LSTM, the model worked perfectly with an autoencoder, but when I try to use a Variational autoencoder by adding the mean and deviation layer and ...
10
votes
1answer
7k views
What is “posterior collapse” phenomenon?
I was going through this paper on Towards Text Generation with Adversarially Learned
Neural Outlines and it states why the VAEs are hard to train for text generation due to this problem. The paper ...
