Introduction to Generative AI

What is Generative AI?

Definition and types of generative models (e.g., GANs, VAEs, Transformer- based models)

Applications of generative AI (text generation, image synthesis, style transfer, etc.)

Types of Generative Models

Generative Adversarial Networks (GANs)

Variational Autoencoders (VAEs)

Autoregressive Models (e.g., PixelCNN, Transformer-based models)

Overview of Generative AI in NLP and Vision

Text generation, image generation, and multimodal generation

Generative Models – GANs and VAEs

Generative Adversarial Networks (GANs)

Introduction to GANs and their components: generator and discriminator

Training GANs: loss functions, challenges like mode collapse

Variants of GANs: Conditional GANs, CycleGAN, StyleGAN

Variational Autoencoders (VAEs)

Introduction to VAEs: encoder-decoder architecture

Latent space representations and sampling

Applications of VAEs in image generation

Introduction to Attention Mechanism

Why Attention is important for Generative Models

Key, Query, and Value vectors

Transformer Architecture

Encoder-Decoder architecture

Multi-head attention, self-attention

Positional Encoding and its importance

Attention in NLP and Image Generation

Language Models and Pretrained Models

Introduction to Pretrained Models in NLP

What are pretrained language models?

Transfer learning in NLP

GPT (Generative Pretrained Transformers)

GPT-1, GPT-2, and GPT-3

How GPT models generate human-like text

Use cases: text completion, dialogue systems, code generation

BERT (Bidirectional Encoder Representations from Transformers)

Understanding BERT architecture

Fine-tuning BERT for specific tasks (text classification, NER, etc.)

Pretrained BERT and its impact on NLP