Scribd
Upload
3 views5 pages

OS

The document discusses advancements in generative AI, focusing on the use of pre-trained models and deep learning techniques for content creation. It highlights the differences between traditional rule-based AI and deep generative models, such as Variational Autoencoders (VAEs), which learn from data patterns without fixed rules. Additionally, it outlines the structure of generative models, including the roles of generators and discriminators in producing and evaluating content.

Uploaded by

saloni.22311814
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
3 views5 pages

OS

The document discusses advancements in generative AI, focusing on the use of pre-trained models and deep learning techniques for content creation. It highlights the differences between traditional rule-based AI and deep generative models, such as Variational Autoencoders (VAEs), which learn from data patterns without fixed rules. Additionally, it outlines the structure of generative models, including the roles of generators and discriminators in producing and evaluating content.

Uploaded by

saloni.22311814
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

&

-
GEN AI UNIT 01
Models create
&
Algos to

New content data


.
E distributi
ou3
NEW CHANGES ?
-

pre-trained models.
-

large
-
same archfor many tasks.
[text > linage > Audio]
- -

audio
synthesised
-

- Generate content
based on .
partial prompts

RULE
BASED Al A
ML
↓
BASED Al 3 OfAl
↓
DEEP LEARNING BASED
↓
GEN-Al & FOUNDATIONAL
MODELS .

PROBABILISTICODELS
-
model learns from patterns
No fixed rules .
-

& How likely


will something
& 0240 .

- --

DEEP GENERATIVE
MODELS
based
learn
probability
description of data .
VAES Variationo
Auto encoder
3
- .

I -
Encodes Decoder
-

compresses -

Expands
data data
-

Img
↓
.
Z

No
No .
S ling
Flow

Image >
-
Encoder >
-
latent
space
Decoder <)
image 1
Small
hidden
Latent space ? -
Space
& stores imp info only 3

NORMAL At VS VAE
Y
↓
stores fixed stores
values
probabilities
Advantages
-
Y
-

smooth gen How to


compress
-

Mix 2 data ?
imag
-

Generate new
easily
data
stable trai by
sampling
-

ing
-

-
2 parts .

--
Generator Discriminator
create Decide ing
fake
- -

images real/fake .

You might also like