Generative Models

• Conditional Generative Models

  • Training models conditioned on specific variables (e.g., text-to-image generation).

• Diffusion Models

  • Study models like DALL-E 2 and Stable Diffusion for image synthesis.

• Text-to-Image and Image-to-Text Generation

  • Leveraging models like CLIP and DALLE for multimodal tasks.

• Energy-Based Models (EBMs)

  • Understand their application in GenAI tasks for density estimation and generation.

• Latent Variable Models

  • Techniques for encoding data in latent spaces for generation.

• Style Transfer and Image-to-Image Translation

  • Generative models for artistic applications.

• Fine-Tuning Generative Models

  • Tailoring generative models to niche applications with limited data.

• Cross-Attention Mechanisms in Generative Models

  • How cross-attention helps in aligning different modalities (text, image, etc.).