Ch.00

Advanced DL: Large Models and Generative AI Paradigm

Advanced Deep Learning (Ch.00) is the entry point that connects “why models got so large” with “how generative AI systems actually work.” We go beyond learning representations from data: how large Transformers build contextual understanding, predict the next token, and then how we align, control, and deploy those models for real users.

An advanced roadmap toward large generative models

This roadmap gradually fills from Ch01 onward, showing how each chapter contributes to the full system.

What you will learn in Ch01~Ch26

Ch.01
Transformer 1: Self-Attention and Parallelization
Ch.02
Transformer: Positional Encoding and Feed-Forward
Ch.03
Transformer Lineage: Encoder (BERT) vs Decoder (GPT)
Ch.04
Attention Optimization: FlashAttention and Sparse Attention
Ch.05
Vision Transformer (ViT) and Image Patches
Ch.06
Self-Supervised Learning
Ch.07
Prompt Engineering and In-Context Learning
Ch.08
PEFT 1: PEFT and LoRA
Ch.09
PEFT 2: QLoRA and Quantization Tuning
Ch.10
Alignment and RLHF
Ch.11
DPO: Alignment without Reinforcement Learning
Ch.12
RAG: Hallucination Control Architecture
Ch.13
LLM Agents and Tool Use
Ch.14
GNN and Message Passing
Ch.15
XAI in Deep Learning: Grad-CAM
Ch.16
Autoencoder and Unsupervised Dimensionality Reduction
Ch.17
VAE: Probability-Based Generative Space
Ch.18
GAN Basics
Ch.19
Conditional GAN (cGAN) and Applications
Ch.20
Diffusion Model 1: Forward and Reverse Process
Ch.21
Diffusion Model 2: Latent Diffusion
Ch.22
Vision-Language Model and CLIP
Ch.23
Speech-to-Text and Audio Processing
Ch.24
Model Compression and Knowledge Distillation
Ch.25
Inference Optimization and Deployment
Ch.26
Advanced DL Summary: AI Architecture and Future

What is Advanced DL? (Generative AI system view)

p(x_t\mid x_{<t})

A practical way to understand generative AI is to split it into stages: pretraining (broad knowledge), instruction / SFT (follow user intent), and alignment (preference, safety, and reduced hallucinations).

The backbone is mostly Transformers . Self-attention creates “token-to-token” context, and feed-forward + normalization layers refine it so the model stays consistent even with long contexts.

Bigger models can improve capability, but they also make training less stable and dramatically increase cost. Advanced DL therefore focuses on more than accuracy: training stability, efficiency (compute/memory), and reproducibility .

In the real world, generative AI is judged by trust: truthfulness, safety, and reliability. Achieving that requires alignment, evaluation, and control mechanisms.

Finally, deployment constraints (latency, cost, server limits) matter. So advanced DL continues from training to inference optimization, compression, and serving strategies .

In production, systems usually follow a pipeline like `text/image -> tokenization -> context window -> Transformer -> decoding (greedy/beam/sample)`. Decoding strategy and prompt design strongly affect output quality.

Alignment and control can be done in multiple ways. For example, RLHF / DPO uses preferences to improve the model, and RAG retrieves external knowledge to ground answers.

From a product perspective, tool use, caching/batching, and optimization such as quantization or knowledge distillation are part of the whole stack. The same base model can feel very different depending on how you run it.

Topic Next-token prediction Role in an AI system Builds general language ability Connection concept in this course probabilistic generation, representation learning Topic Instruction / SFT Role in an AI system Makes responses follow intent Connection concept in this course data format, fine-tuning Topic Alignment Role in an AI system Controls preferences, safety, truthfulness Connection concept in this course preference learning, reward model Topic RAG / grounded generation Role in an AI system Reduces ungrounded claims Connection concept in this course retrieval, embeddings, context assembly Topic Inference optimization Role in an AI system Cuts latency and cost Connection concept in this course quantization, caching, distillation Topic Role in an AI system Connection concept in this course Next-token prediction Builds general language ability probabilistic generation, representation learning Instruction / SFT Makes responses follow intent data format, fine-tuning Alignment Controls preferences, safety, truthfulness preference learning, reward model RAG / grounded generation Reduces ungrounded claims retrieval, embeddings, context assembly Inference optimization Cuts latency and cost quantization, caching, distillation

What is Advanced DL? (Generative AI system view)

p(x_t\mid x_{<t})

A practical way to understand generative AI is to split it into stages: pretraining (broad knowledge), instruction / SFT (follow user intent), and alignment (preference, safety, and reduced hallucinations).

The backbone is mostly Transformers . Self-attention creates “token-to-token” context, and feed-forward + normalization layers refine it so the model stays consistent even with long contexts.

Bigger models can improve capability, but they also make training less stable and dramatically increase cost. Advanced DL therefore focuses on more than accuracy: training stability, efficiency (compute/memory), and reproducibility .

In the real world, generative AI is judged by trust: truthfulness, safety, and reliability. Achieving that requires alignment, evaluation, and control mechanisms.

Finally, deployment constraints (latency, cost, server limits) matter. So advanced DL continues from training to inference optimization, compression, and serving strategies .

In production, systems usually follow a pipeline like `text/image -> tokenization -> context window -> Transformer -> decoding (greedy/beam/sample)`. Decoding strategy and prompt design strongly affect output quality.

Alignment and control can be done in multiple ways. For example, RLHF / DPO uses preferences to improve the model, and RAG retrieves external knowledge to ground answers.

From a product perspective, tool use, caching/batching, and optimization such as quantization or knowledge distillation are part of the whole stack. The same base model can feel very different depending on how you run it.

Topic Next-token prediction Role in an AI system Builds general language ability Connection concept in this course probabilistic generation, representation learning Topic Instruction / SFT Role in an AI system Makes responses follow intent Connection concept in this course data format, fine-tuning Topic Alignment Role in an AI system Controls preferences, safety, truthfulness Connection concept in this course preference learning, reward model Topic RAG / grounded generation Role in an AI system Reduces ungrounded claims Connection concept in this course retrieval, embeddings, context assembly Topic Inference optimization Role in an AI system Cuts latency and cost Connection concept in this course quantization, caching, distillation Topic Role in an AI system Connection concept in this course Next-token prediction Builds general language ability probabilistic generation, representation learning Instruction / SFT Makes responses follow intent data format, fine-tuning Alignment Controls preferences, safety, truthfulness preference learning, reward model RAG / grounded generation Reduces ungrounded claims retrieval, embeddings, context assembly Inference optimization Cuts latency and cost quantization, caching, distillation

Topic	Role in an AI system	Connection concept in this course
Next-token prediction	Builds general language ability	probabilistic generation, representation learning
Instruction / SFT	Makes responses follow intent	data format, fine-tuning
Alignment	Controls preferences, safety, truthfulness	preference learning, reward model
RAG / grounded generation	Reduces ungrounded claims	retrieval, embeddings, context assembly
Inference optimization	Cuts latency and cost	quantization, caching, distillation