Generative Models

Generative Models:

A Generative model is a model for generating all values for a phenomenon, both those that can be observed in the world and "target" variables that can only be computed from those observed. Discriminative models infer outputs based on inputs, while generative models generate both inputs and outputs, typically given some hidden parameters.

Suppose the input data is x and the set of labels for x is y. A generative model learns the joint probability distribution p(x,y) while a discriminative model learns the conditional probability distribution p(y|x) “probability of y given x”.

Examples: Generative: Naive Bayes, Latent Dirichlet Allocation, Probabilistic context-free grammar, Hidden Markov Models, Generative Adversarial Network

Discriminative: Logistic regression, Support Vector Machines, Maximum Entropy Markov Model, Conditional Random Fields, Neural Networks

Generative Models don't just spit out the classification but create whole new image or sound. They can be used with images for:

1. Improve photos (colorization, noise removal, increase resolution)
2. Create Art
3. Find and segment/localize objects

AutoEncoders

Basics