AI GlossaryDiffusion ModelENTRY — Generative AI
[Generative AI]
Diffusion Model.
Generative models that learn to create data by reversing a gradual noising process.
In-depth explanation
01Diffusion models add noise to data in small steps, then learn to reverse this process, generating data from pure noise. They've achieved state-of-the-art image generation quality, surpassing GANs. Stable Diffusion, DALL-E, and Midjourney use diffusion. The approach is more stable to train than GANs and offers better sample diversity.
Examples
02EX. 01
Stable Diffusion
EX. 02
DALL-E 2
EX. 03
Midjourney
More in Generative AI
0301Generative Adversarial Network (GAN)Two neural networks competing against each other to generate realistic synthetic data.02GPTGenerative Pre-trained Transformer, a family of large language models trained to generate text.03HallucinationWhen AI models generate plausible-sounding but factually incorrect or fabricated information.04Large Language Model (LLM)AI models trained on vast text data that can generate and understand human-like text.05Prompt EngineeringThe practice of crafting effective inputs to get desired outputs from AI models.06RAG (Retrieval-Augmented Generation)Combining retrieval systems with language models to generate responses grounded in external knowledge.
[NEXT] — APPLY THE CONCEPT
Master Diffusion Model.
Learn how to apply this concept with hands-on projects in our comprehensive AI programs.