Recap of Deep Neural Networks, Autoencoders, Variational Autoencoders (VAEs), attention mechanisms, and fundamentals of generative modelling. Introduction to Generative Adversarial Networks (GANs): generator and discriminator architectures, adversarial training, loss functions, and optimization challenges. Advanced GAN architectures: Deep Convolutional GANs (DCGAN), Conditional GANs, Pix2Pix, CycleGAN, and StyleGAN. Applications of GANs in image synthesis, data augmentation, and content generation. Challenges in GANs including mode collapse, bias, fairness, and privacy preservation.
Introduction to diffusion models: principles of forward noising and reverse denoising processes. Denoising Diffusion Probabilistic Models (DDPMs), score-based generative models, latent diffusion models, and Stable Diffusion. Applications of diffusion models in text-to-image generation, image editing, inpainting, and super-resolution. Comparative analysis of VAEs, GANs, and diffusion models. Evaluation techniques for generated images including Fréchet Inception Distance (FID), diversity, realism assessment, and human evaluation methods.
Principles of Transformer architecture: self-attention mechanisms, positional encoding, tokenization, embeddings, and transformer variants including encoder-only, decoder-only, and encoder-decoder models. Overview of BERT and GPT-based models, pretraining and fine-tuning strategies. Prompt engineering, instruction tuning, supervised fine-tuning, parameter-efficient fine-tuning techniques including LoRA and QLoRA, quantization, small language models, Reinforcement Learning from Human Feedback (RLHF), Direct Preference Optimization (DPO), and LLM alignment. Applications of Large Language Models in text generation, summarization, translation, question answering, sentiment analysis, chatbots, healthcare, and code generation. Retrieval-Augmented Generation (RAG), embeddings, vector databases, semantic search, LLM evaluation, and Introduction to Agentic AI: LLM agents, tool usage, memory, planning, workflow orchestration, and multi-agent systems.