Knowledge Graph — Coursera Notes › Academic disciplines › Information Technology / Computer Science › Artificial Intelligence › Generative AI

Generative AI Models

concept · part of Generative AI

Machine learning architectures designed to create new content (text, images, audio, video) by learning patterns from vast datasets. Key types include:

• Transformers: excel at sequential data (e.g., text) using self-attention mechanisms for human-like language generation.
• Generative Adversarial Networks (GANs): competitive framework of generator and discriminator to create realistic outputs.
• Variational Autoencoders (VAEs): encode data into latent space and decode for variations; ideal for image generation or anomaly detection.
• Diffusion models: reverse noise to generate high-quality images; perfect for design applications.

Training involves large, high-quality datasets, unsupervised learning, gradient descent optimization, and transfer learning for task adaptation.

• Content creation: LLMs streamline writing workflows; transformers and diffusion models enable artists to explore new creative directions.
• Medical imaging: VAEs support anomaly detection for earlier diagnoses.
• Gaming: transformers enhance dynamic and personalized content.
• Automation: automate repetitive tasks, generate creative content, solve complex challenges.

Select the model that best fits your use case: LLMs for text generation, GANs for realistic image creation, Diffusion models for high-quality visual output like animation. Balance creativity with control by fine-tuning parameters or adding constraints to guide outputs while maintaining flexibility.

Inside Generative AI Models (3)

• Diffusion models — A generative AI model that reverses noise to generate high-quality images.
• Generative Adversarial Networks (GANs) — A generative AI model using a competitive framework of a generator and a discriminator to create realistic outputs.
• Variational Autoencoders (VAEs) — A generative AI model that encodes data into a latent space and decodes it for variations.

This is the text view of an interactive 3D knowledge graph — open this page with JavaScript enabled to explore it visually.