Diffusion model concept in Prompt Engineering / GenAI - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is a diffusion model in simple terms?

A diffusion model is a type of AI that learns to create data by slowly adding noise to it and then learning how to remove that noise step-by-step to get back the original data.

beginner

Why do diffusion models add noise to data during training?

They add noise to teach the model how to reverse the process, so it can start from random noise and create new, realistic data by removing noise gradually.

beginner

What is the main goal of the reverse diffusion process?

The goal is to start with pure noise and step-by-step remove noise to generate new data that looks like the original training examples.

beginner

How is a diffusion model similar to making a blurry photo clear again?

Just like making a blurry photo clear by removing blur step-by-step, a diffusion model removes noise step-by-step to create clear, realistic data.

beginner

Name one popular use of diffusion models in AI today.

Diffusion models are popular for generating images, like creating art or photos from text descriptions.

What does a diffusion model learn to do?

A Remove noise from data step-by-step

B Add noise to data only

C Classify images into categories

D Translate text from one language to another

Diffusion models learn to remove noise gradually to generate new data.

Why is noise added during training in diffusion models?

A To make data harder to understand

B To speed up training

C To reduce the size of the data

D To teach the model how to reverse noise addition

Noise is added so the model learns how to remove it step-by-step.

What is the starting point for generating new data in a diffusion model?

A Clean original data

B Random noise

C A labeled dataset

D A trained classifier

Generation starts from random noise that the model cleans up.

Which of these is a common application of diffusion models?

A Predicting stock prices

B Sorting numbers

C Image generation from text

D Detecting spam emails

Diffusion models are widely used to generate images from text prompts.

How does the reverse diffusion process work?

A By removing noise step-by-step

B By adding more noise each step

C By copying data exactly

D By randomly guessing data

Reverse diffusion removes noise gradually to create new data.

Explain in your own words how a diffusion model generates new data.

Think about how the model learns by reversing noise addition.

Describe a real-life analogy that helps you understand the diffusion model concept.

Imagine cleaning a dirty window or fixing a blurry picture.