Scribd
Upload
4 views

class notes astronomy 3 of 5

Generative AI encompasses algorithms that create new content by learning the distribution of existing data, contrasting with discriminative models that classify data. Key models include Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and autoregressive models like GPT-3, each with distinct applications in image, text, music, and code generation. The document outlines the fundamental differences between generative and discriminative models, emphasizing their respective methodologies and use cases.

Uploaded by

secretgalatea
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
4 views

class notes astronomy 3 of 5

Generative AI encompasses algorithms that create new content by learning the distribution of existing data, contrasting with discriminative models that classify data. Key models include Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and autoregressive models like GPT-3, each with distinct applications in image, text, music, and code generation. The document outlines the fundamental differences between generative and discriminative models, emphasizing their respective methodologies and use cases.

Uploaded by

secretgalatea
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 2

Generative AI 101: Introduction to Generative Models

Class Notes – Week 2: Fundamentals of Generative AI


Instructor: Dr. Lucas Jenkins
Date: February 11, 2025

Key Concepts:
Generative AI refers to algorithms that can generate new content, such as images,
text, music, or even code, that mimics existing data. Unlike discriminative models
(which classify data), generative models learn the distribution of the data itself
and can produce new instances of it.

Applications of Generative AI:

Image Generation (e.g., DALL·E, GANs)


Text Generation (e.g., GPT-3, T5)
Music Composition (e.g., OpenAI’s Jukedeck, Google's Magenta)
Video Generation (e.g., Deepfake technology)
Code Generation (e.g., GitHub Copilot, Codex)
1. Generative vs. Discriminative Models
Generative Models:
These models learn how the data is generated, allowing them to create new, similar

𝑃
data. They model the joint probability distribution

𝑋
(

𝑌
,

𝑋
P(X,Y), where

𝑌
X is the data and

Y is the label.

Example: Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs).


Discriminative Models:
These models focus on learning the boundaries between different classes of data.

𝑃
They model the conditional probability

𝑌
(

𝑋
∣

𝑌
P(Y∣X), where

𝑋
Y is the label and

X is the feature.

Example: Logistic Regression, Support Vector Machines (SVMs), Random Forests.


2. Key Generative AI Models
Generative Adversarial Networks (GANs):

GANs consist of two neural networks: a generator and a discriminator.


Generator: Tries to generate data that is indistinguishable from real data.
Discriminator: Tries to differentiate between real and fake data.
The two networks are trained together in a game-like setup where the generator gets
better at generating data, and the discriminator gets better at spotting fake data.
Applications: Image synthesis, super-resolution, style transfer, etc.
Variational Autoencoders (VAEs):

VAEs are probabilistic generative models that learn an encoding of data and then
decode it back into the original data distribution.
The encoder maps input data to a latent space, while the decoder reconstructs the
data from this latent representation.
Unlike GANs, VAEs have a more structured latent space, which helps in controlling
the generation process.
Applications: Image generation, anomaly detection, and semi-supervised learning.
Autoregressive Models:

These models generate data sequentially, one step at a time, based on the preceding
steps.
Example: GPT-3 (text generation), WaveNet (audio generation).

𝑃
They work by modeling the conditional probability

𝑥
(

𝑥
∣

𝑥
,

2
,
.
.
.

𝑥
,

𝑡
−
1
)
P(x
t

∣x
1

,x
2

,...,x
t−1

), predicting each data point in the sequence given the previous ones.

You might also like