Phase 02 — Learning and Models

This phase turns objectives and computation into learned representations and decisions. Study each architecture as an information path, an inductive bias, and a cost model—not a brand name.

Mental model

A model architecture restricts how information and gradients can move. Learning selects parameters within that structure; data, objective, and optimization decide which behavior emerges.

Roadmap through the branches

• Deep Learning
• Reinforcement Learning
• Model Architectures
• Language and Foundation Models
• Vision, Audio and Multimodal AI

Exit criteria

You can trace a forward computation, name the learned parameters and objective, explain the update signal, compare architecture tradeoffs, and identify what scale changes.

Connects to: Phase 01 — Foundations · Phase 03 — Training and Inference

Core sources

• Deep Learning — neural learning foundations.
• Attention Is All You Need — transformer architecture.
• Reinforcement Learning: An Introduction — sequential decision-making foundations.