Reference · Glossary
Deep learning
Last updated
Perception and generation problems with lots of examples: images, speech, language, recommendation ranking. LLMs and diffusion models are deep learning systems.
#When to use
Perception and generation problems with lots of examples: images, speech, language, recommendation ranking. LLMs and diffusion models are deep learning systems.
#When not to
Tiny tabular datasets where a simple model (or rules) already works and you need easy explanations. Deep nets need data, compute, and careful evals.
#Quality checklist
- Train/validation/test splits that match real use
- Baseline simpler models before going deep
- Monitor overfitting and distribution shift
- Document data provenance and failure cases
#Example
Task: classify support tickets into 8 topics.
Start: bag-of-words logistic regression baseline.
Only move to a fine-tuned encoder if F1 gains justify the ops cost.#Learn next
- Lesson: `deep-learning-basics`
- Related: neural network, transformer