Editing Meta-Learning (section)

== <span style="color: #FFFFFF;">Remembering</span> ==
* '''Meta-learning''' — Learning to learn; training models to adapt quickly to new tasks using experience across many tasks.
* '''Meta-learner''' — The higher-level model that learns across tasks and produces or adapts base learners.
* '''Base learner''' — The model that is applied to individual tasks; updated by the meta-learner.
* '''Support set''' — The small labeled dataset provided at test time to adapt to a new task (analogous to training data for the base learner).
* '''Query set''' — The test examples for the new task on which performance is evaluated after adaptation.
* '''N-way K-shot learning''' — A meta-learning task setting: N classes, K labeled examples per class in the support set.
* '''Episode''' — One meta-learning training iteration, consisting of a sampled task with its support and query sets.
* '''MAML (Model-Agnostic Meta-Learning)''' — A gradient-based meta-learning algorithm that finds initialization parameters enabling rapid fine-tuning on new tasks.
* '''Prototypical Networks''' — A metric-based meta-learning approach that classifies by distance to class prototype embeddings.
* '''Matching Networks''' — A metric-based approach using an attention mechanism over support set embeddings.
* '''Meta-SGD''' — An extension of MAML that also meta-learns per-parameter learning rates.
* '''In-context learning''' — The emergent ability of large LLMs to learn new tasks from examples provided in the prompt, without gradient updates.
* '''Hyperparameter optimization (HPO)''' — Automatically finding optimal hyperparameters; meta-learning approaches (BOHB, SMAC) use experience across runs.
</div>

<div style="background-color: #006400; color: #FFFFFF; padding: 20px; border-radius: 8px; margin-bottom: 15px;">