Generative Models

A generative model combining autoregressive and flow-based methods for flexible data distribution learning.

Beta-VAE is a type of variational autoencoder that focuses on disentangling learned representations by adjusting a hyperparameter, beta.

A Conditional Variational Autoencoder (CVAE) is a type of neural network that generates data conditioned on specific input labels.

CycleGAN is a type of neural network that enables image-to-image translation without paired examples.

Deep Generative Models are AI systems that learn to create new data samples similar to existing data.

FID Score measures the quality of generated images by comparing them to real images.

Flow-Based Generative Models use invertible transformations to generate high-dimensional data from simpler distributions.

GAN Collapse refers to a phenomenon where a Generative Adversarial Network fails to generate diverse outputs, often producing similar results.

GAN Inversion refers to the process of mapping real images back into the latent space of a Generative Adversarial Network.

GAN Space refers to the latent space of Generative Adversarial Networks, where different points correspond to unique generated outputs.

Generative Flow Networks are AI models that generate data by learning complex distributions through continuous transformations.

GIFA Loss is a metric used to evaluate generative models based on their ability to generate realistic samples.

The Glow Model is a generative model used for creating complex data distributions, particularly in AI and deep learning.

Goodfellow GAN is a type of generative adversarial network that generates realistic data through adversarial training.

Gradient Penalty is a regularization term used in machine learning to improve model stability and performance.

A Helmholtz Machine is a type of generative model that learns to represent data distributions.

Masked Autoregressive Flow is a neural network technique for generating complex data distributions using autoregressive models.

Mode collapse occurs when a generative model produces limited diversity in outputs, focusing on a few patterns.

PixelCNN is a deep learning model for generating images pixel by pixel using convolutional neural networks.

RealNVP is a type of deep learning model used for generative tasks, enabling efficient data sampling and density estimation.

A score-based generative model generates new data by learning the score function of a probability distribution.

Stable Diffusion is a deep learning model for generating images from text prompts.

A Variational Autoencoder (VAE) is a type of neural network that generates new data similar to a training dataset.