Flow-based generative model

A flow-based generative model is a <a href="/facts/Generative_model/JUgExNIP">generative model</a> used in <a href="/facts/Machine_learning/e0w0XJTu">machine learning</a> that explicitly models a <a href="/facts/Probability_distribution/EpsKKVRu">probability distribution</a> by leveraging normalizing flow, which is a statistical method using the <a href="/facts/Probability_density_function/zvfybna4">change-of-variable</a> law of probabilities to transform a simple distribution into a complex one.
The direct modeling of likelihood provides many advantages. For example, the negative log-likelihood can be directly computed and minimized as the <a href="/facts/Loss_function/xv5ozuhl">loss function</a>. Additionally, novel samples can be generated by sampling from the initial distribution, and applying the flow transformation.
In contrast, many alternative generative modeling methods such as <a href="/facts/Variational_autoencoder/I5wBbYpR">variational autoencoder (VAE)</a> and <a href="/facts/Generative_adversarial_network/SGNuGrlo">generative adversarial network</a> do not explicitly represent the likelihood function.

Flow-based generative model open-in-new

Flow-based generative model