Questions tagged [generative-model]

For questions related to the concept of generative machine learning models, such as the Restricted Boltzmann Machine (RBM), the Variational Autoencoder (VAE), and the Generative Adversarial Network (GAN).

Filter by
Sorted by
Tagged with
0
votes
1answer
22 views

Is the range of inception score flexible or bounded based on number of classes?

Inception score is used to evaluate the generative models. It is a score given based on quality and diversity of images generated. I have doubt about the range of inception score because of the reason ...
3
votes
1answer
56 views

How to define loss function for Discriminator in GANs?

To train the discriminator network in GANs we set the label for the true samples as $1$ and $0$ for fake ones. Then we use binary cross-entropy loss for training. Since we set the label $1$ for true ...
2
votes
1answer
30 views

What is the fundamental difference between the synthesis task and sampling task?

Among the list of tasks in machine learning, synthesis and sampling is one of the key task. Consider the following explanation regarding synthesis and sampling task from Chapter 5: Machine Learning ...
0
votes
0answers
11 views

Are ranking models considered discriminative?

I'm developing a model that ranks entries based on cosine similarity to a query. Since it doesn't actually define a boundary between x and y I initially believed that such ranking models are not ...
4
votes
1answer
424 views

What is the input for the prior model of VQ-VAE?

I'm trying to implement the VQ-VAE model. In there, a continuous variable $x$ is encoded in an array $z$ of discrete latent variables $z_i$ that are mapped each to an embedding vector $e_i$. These ...
2
votes
1answer
42 views

What is the state of the art in melody generation?

Generative Adversarial Networks can generate realistic photos of people, such as thispersondoesnotexist.com. I wonder whether one can train an artificial intelligence on a batch of plain solo melodies ...
0
votes
0answers
16 views

Types of decoder parametrizations in VAE for continuous data

I'm wondering what are the different choices of parametrizations available for the decoder in a variational autoencoder. If the data is discrete, you can just output probabilities for each class, so ...
1
vote
1answer
40 views

Is image generation not existent before generative adversarial networks?

Although the GAN is widely used due to its capability, there were generative models before the GAN which are based on probabilistic graphical models such as Bayesian networks, Markov networks, etc. It ...
0
votes
1answer
46 views

Is it possible to use deep learning to generate a 2D image from a few numerical values?

Is it possible to train a DL model that will generate a full resolution 2D image based on few numbers describing this image and what type of model or architecture would that be? What I want to ...
1
vote
1answer
908 views

What is the purpose of the GAN?

The Generative Adversarial Network (GAN) is composed of a generator $G$ and a discriminator $D$. How do these two components interact? What is the intuition behind the GAN, its purpose, and how it is ...
0
votes
2answers
112 views

What is the right way to train a generator in a GAN?

I am not fully understanding how to train a GAN's generator. I have a few questions below, but let me first describe what I am doing. I am using the MNIST dataset. I generate a batch of random images ...
2
votes
3answers
2k views

Confusing on GAN loss function

I was trying to understand the loss function of GANs, while I found a little mis-match between different papers. This is the screen-shot from the original paper of Goodfellow at https://arxiv.org/...
1
vote
1answer
161 views

Optimum Discriminator for label smoothed GAN

I was reading the paper called Improved Techniques for Training GANs. And, in the one-sided label smoothing part, they said that optimum discriminator with label smoothing is $$ D^*(x)=\frac{\alpha \...
1
vote
1answer
25 views

Book(s) on generative models

Generative models in artificial intelligence span from simple models like Naive Bayes to the advanced deep generative models like current day GANs. This question is not about coding and involves only ...
1
vote
1answer
36 views

Why is this variable in equation 2 of the SQAIR paper a random vector of $n$ ones followed by a zero?

I've been reading the SQAIR paper lately, and the mathematics involved seems a bit complicated. Some background, about the paper: SQAIR stands for Sequential Attend, Infer, Repeat - the paper does ...
0
votes
0answers
13 views

What approaches are there to generate complex structures like syntactic trees?

What approaches are there to generate and evaluate complex structures like, let's say, syntactically correct code? I know the approach of Genetic Programming (GP) as a type of Evolutionary Algorithm, ...
0
votes
0answers
19 views

GAN - relation of input size and hidden layer size

I'm adapting a GAN described here used for generating binary output. It's trained on binarized MNIST data, with a size of 28x28 so 784 values. I want to adapt it to train on and generate 1D vectors ...
1
vote
0answers
19 views

Necessity of likelihood in training energy-based models

Lately, I've been getting into energy-based models (EBMs) through some of Yann LeCun's recent talks, where he advocates the use of non-normalized models because it allows for more flexibility in the ...
11
votes
3answers
429 views

Can some one help me understand this paragraph from Nvidia's progressive gan paper?

Furthermore, we observe that mode collapses traditionally plaguing GANs tend to happen very quickly, over the course of a dozen minibatches. Commonly they start when the discriminator overshoots, ...
0
votes
1answer
62 views

Why is it easier to construct adversarial examples relative to training neural networks?

I was having looking at this lecture by Ian Goodfellow and my doubt is around 18:00 timestamp where he explains generation of adversarial examples using FGSM. He mentions that the there is a linear ...
4
votes
0answers
40 views

Are generative models actually used in practice for industrial drug design?

I just finished reading this paper MoFlow: An Invertible Flow Model for Generating Molecular Graphs. The paper, which is about generating molecular graphs with certain chemical properties improved the ...
0
votes
1answer
64 views

Decision boundary figure in Least square GAN paper

I currently reading Least Square GAN paper. But, I cannot interpret the one of the its figures. . Explanation of the figure goes like this: Figure 1: Illustration of different behaviors of two loss ...
12
votes
1answer
1k views

Understanding notation of Goodfellow's GAN objective function

What is the meaning of $V(D,G)$? How do we get these expectation parts? I was trying to understand it following this article: Understanding Generative Adversarial Networks (D.Seita), but, after many ...
0
votes
0answers
63 views

Are there any good tutorials on using continuous normalizing flows (with PyTorch)?

I just have a very general question. Are there any good tutorials on using continuous normalizing flows? I'd say I have a decent understanding of normalizing flows, but not their continuous variant. I'...
0
votes
0answers
14 views

Why do the inception score and the Fréchet inception distance use the inception network and not another network?

So I was researching about the evaluation of GANs and found these two metrics which seem to be the most popular. I understand that the main ideia is to apply the data to a pre-trained network in order ...
2
votes
1answer
67 views

Would it be possible to determine the dataset a neural network was trained on?

Let's say we have a neural network that was trained with a dataset $D$ to solve some task. Would it be possible to "reverse-engineer" this neural network and get a vague idea of the dataset $...
7
votes
2answers
418 views

What are the best machine learning models for music composition?

What are the best machine learning models that have been used to compose music? Are there some good research papers (or books) on this topic out there? I would say, if I use a neural network, I would ...
2
votes
2answers
106 views

What are the possible social consequences of training neural networks with artificially generated data?

Machine learning models and, in particular, neural networks are trained with data often collected from the real world, such as images of real people. Meanwhile, neural networks (such as GANs) are also ...
0
votes
0answers
11 views

Parametrizing non-analytical functions using generative models

My questions centers around what method is best to use parametrize a response function for an experiment. We are currently using ab initio simulation to model our experiment's response. Unfortunately, ...
3
votes
2answers
149 views

Mathematical Analysis of the Loss function of GAN

I was pondering on loss function of GAN and following thing turned out \begin{aligned} L(D, G) & = \mathbb{E}_{x \sim p_{r}(x)} [\log D(x)] + \mathbb{E}_{x \sim p_g(x)} [\log(1 - D(x)] \\ & =...
0
votes
0answers
66 views

How to compare multiple one-class variational autoencoders?

I have trained multiple one-class vanilla variational autoencoders that each learn the distribution of one class and have the same architecture. The classes are mostly discrete, but there are several ...
0
votes
0answers
21 views

Network structure of generative model for classification

I'm trying to model a generative model for classification problem, especially aiming to solve an imbalanced data problem. However, I couldn't get intuitive understanding for generative classifier in ...
0
votes
0answers
9 views

Given a system state, generate a sequence of state changes that lead to it

These systems are discrete and their state changes are rule based. Example: Given a chess position, generate a series of moves that will lead to it (there may be many, one, or none, but I only need ...
2
votes
0answers
103 views

Generating fake faces containing specific features with GANs

I'm trying to understand how DeepFakes are generated and so far I understood that they're mostly generated through the usage of GANs and autoencoders. The autoencoders part is understandable, but what ...
1
vote
0answers
58 views

How does NN follows law of energy conservation?

Communication requires energy, and using energy requires communication. According to Shannon, the entropy value of a piece of information provides an absolute limit on the shortest possible average ...
20
votes
3answers
25k views

How can we process the data from both the true distribution and the generator?

I'm struggling to understand the GAN loss function as provided in Understanding Generative Adversarial Networks (a blog post written by Daniel Seita). In the standard cross-entropy loss, we have an ...
1
vote
1answer
50 views

Is GAIL applicable if the expert's trajectories are for the same task but are in a different environment?

Is the GAIL applicable if the expert's trajectories (sample data) are for the same task but are in a different environment (modified but will not be completely different)? My gut feeling is, yes, ...
1
vote
1answer
28 views

Adding voices to voice synthesis corpuses

If one uses one of the open source implementations of the WaveNet generative speech synthesis design, such as https://r9y9.github.io/wavenet_vocoder/, and trains using something like the CMU's arctic ...
1
vote
1answer
776 views

Why do we need Upsampling and Downsampling in Progressive Growing of Gans

I was working recently on Progressive Growing of GANs (aka PGGANs). I have implemented the whole architecture, but the problem that was ticking my mind is that in simple GANs, like DCGAN, PIX2PIX, we ...
1
vote
1answer
96 views

Is it legal to license and sell the output of a neural network that was trained on data that you don't own the license to?

Is it legal to license and sell the output of a neural network that was trained on data that you don't own the license to? For example, suppose you trained WaveNet on a collection of popular music. ...
0
votes
2answers
137 views
5
votes
1answer
3k views

Are deep learning models suitable for training with sparse data?

I am training a generative adversarial network (GAN) to generate images given edge histogram descriptor (EHD) features of the image. The EHD features are themselves sparse (meaning they contain a lot ...
1
vote
1answer
62 views

What is meant by degrees of freedom of latent variables?

...Designing such a likelihood function is typically challenging; however, we observe that features like spectrogram are effective when latent variables have limited degrees of freedom. This motivates ...
1
vote
0answers
33 views

Concrete example of how transposed convolutions are able to *add* features to an image

Say we have a simple gray scale image. If we use a filter which is just the 3x3 identity matrix (or more pointedly the identity matrix but with -1 instead of the 0 entries), it is fairly easy to see ...
0
votes
2answers
62 views

How to combine several chatbots into one?

I'm in the middle of a project in which I want to generate a TV series script (characters answering to each other, scene by scene) using SOTA models, and I need some guidance to simplify my ...
1
vote
0answers
35 views

How do I sample conditionally from deep belief networks?

Deep belief networks (DBNs) are generative models, where, usually, you sample by thermalising the deepest layer (as it's a restricted Boltzmann machine), and then forward propagating a sample towards ...
1
vote
1answer
92 views

Why does the discriminator minimize the cross-entropy while the generator maximize it?

In his original GAN paper Goodfellow gives a game theoretic perspective for GANs: \begin{equation} \underset{G}{\min}\, \underset{D}{\max}\, V\left(D,G \right) = \mathbb{E}_{x\sim\mathit{p}_{\...
2
votes
1answer
87 views

Why do we use $D(x \mid y)$ and not $D(x,y)$ in conditional generative adversarial networks?

In conditional generative adversarial networks (GAN), the objective function (of a two-player minimax game) would be $$\min _{G} \max _{D} V(D, G)=\mathbb{E}_{\boldsymbol{x} \sim p_{\text {data }}(\...
2
votes
0answers
22 views

Why do hypercube latent spaces perform poorer than Gaussian latent spaces in generative neural networks?

I have a quick question regarding the use of different latent spaces to represent a distribution. Why is it that a Gaussian is usually used to represent the latent space of the generative model ...
4
votes
1answer
296 views

Does MMD-VAE solve the problem of blurred images of vanilla VAEs?

I understand that with vanilla VAEs, there are a few reasons justifying the production of blurred out images. The InfoVAE paper describes the case when the decoder is flexible enough to ignore the ...