AI & MLRecommendedAIML-TRF-012
Mixture-of-Experts Layers
Routing each token to a few expert feed-forward networks: more parameters without more compute per token.
- Difficulty
- 4 / 5
- Importance
- 4 / 5
- Study time
- 1–2 h
- Kind of work
- Mixed
Why it matters
The learner can trace token routing through a top-k MoE layer and explain the load-balancing problem.