Neural Networks, Embeddings, and Transformers
The complete concept, practice, assessment, transfer, and source-routing path is available.
Selective Reading Rule
The guide is the primary teacher. Full-book reading is not required; use the module reading guide and linked primary sources only to resolve a named question or deepen a decision.
Competency
Develop intuition from optimization and representations through attention and transformers; distinguish training, inference, adaptation, and retrieval.
Required Evidence
- small trained model
- embedding analysis
- architecture explanation
- controlled ablation
Completion Gate
Explain the underlying mechanism, reproduce the result from a clean environment, evaluate at least one rejected alternative, and defend the important failure modes without relying on generated prose.
Source Policy
Use official specifications, primary research, and curated local book routes selectively. The completed guide must remain the primary teacher.