L9 - Deep Generative Models

Question 1

What is a Generative Model?

Answer 1

A generative model is a type of machine learning model that learns to generate new data that resembles the training data. Instead of just making predictions (like a classifier), a generative model tries to understand the underlying distribution of the data so it can create new, similar examples.

Question 2

What is an Autoencoder?

Answer 2

An unsupervised Neural Network. An autoencoder learns to compress data into a smaller representation (called the latent space) and then reconstruct the original data from that compressed version.

It has two main parts:

Encoder: Compresses the input into a lower-dimensional latent vector.
Decoder: Reconstructs the input from the latent vector.

Question 3

What is a Variational Autoencoder?

Answer 3

A Variational Autoencoder (VAE) is a type of generative model that builds on the basic autoencoder architecture but adds a probabilistic twist. It’s designed not just to reconstruct input data, but to learn a meaningful latent space from which new data can be generated.

Question 4

What is a Generative Adversarial Network (GAN)?

Answer 4

A GAN, or Generative Adversarial Network, is a powerful type of generative model that uses two neural networks — a generator and a discriminator — that compete against each other in a game-like setup. This competition helps the generator learn to produce realistic, high-quality data, such as images, audio, or text.