Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
138 views13 pages

Introduction To Generative Models

This document provides an introduction to generative models. It defines generative models as models that address density estimation to generate new samples from the same probability distribution as the training data. The document discusses supervised and unsupervised learning, and explains that generative models learn the joint probability distribution rather than the conditional probability. It provides examples of applications of generative models such as image generation and translation, speech generation, and sequence generation. Finally, it outlines some common types of generative models including auto-regressive models, variational autoencoders, and generative adversarial networks.

Uploaded by

Prudhvi Vuda
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
138 views13 pages

Introduction To Generative Models

This document provides an introduction to generative models. It defines generative models as models that address density estimation to generate new samples from the same probability distribution as the training data. The document discusses supervised and unsupervised learning, and explains that generative models learn the joint probability distribution rather than the conditional probability. It provides examples of applications of generative models such as image generation and translation, speech generation, and sequence generation. Finally, it outlines some common types of generative models including auto-regressive models, variational autoencoders, and generative adversarial networks.

Uploaded by

Prudhvi Vuda
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

Introduction to Generative Models

Content
● Background

● What is Generative Model?

● Why Generative Models?

● Types of Generative Models


Supervised Learning

● Data available is (X,Y), i.e.,


both data(X) and label(Y) are
available.

● Aim: To learn mapping from X--


>Y

● Example: Classification,
regression, object detection,
semantic segmentation, image
captioning etc.
Unsupervised learning
● Only data(X) is available and
label(Y) is not known

● Aim: Learn some hidden


representation of data

● Example: Clustering,
dimensionality reduction, feature
learning, density estimation, etc.
What is Generative Model?
● Model that addresses the density
estimation of the data, a core
problem in unsupervised learning
● Density estimation: Generate new
samples following the same
probability distribution as that of
given training dataset
What is Generative Model?

● Generative Model learns joint probability distribution, P(X,Y), whereas, the


distributive model learns posterior probability, P(Y|X)

● P(X,Y)=P(Y|X).P(X)
Why Generative Models? (Application)
1. Generates new images, using image to image translation

Image-to-Image Translation with Conditional Adversarial Networks, by Isola et al.


2. Generating speech from text

WaveNet: A generative model for raw audio, by Van Den Oord et al.
3. Generating Sequences

Generating Sequences With Recurrent Neural Networks, by Alex Graves


4. Generating images with super resolution

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial


Network, by Ledig et al.
Applications
● Generate data for training

○ Even unlabelled patient data is hard to obtain- would it be possible to generate medical
images of specific anatomy?
● Image to image translation for medical imaging applications

○ Given a trained model for a specific modality (CT) and pathology – can we transform from
one modality to another

○ E.g. Given CT images and a trained model for diagnosis- Transform MR images of same
anatomy and use the CT trained model for diagnosis
Types of Generative Models
● Auto-regressive models- PixelRNN, PixelCNN
● Latent Variable Model- Variational Autoencoders- VAE
● Implicit density model: Generative Adversarial Networks- GANs
Thank you for your attention!
Questions/comments?

You might also like